diff --git "a/examples/talk-llama/llama-model.cpp" "b/examples/talk-llama/llama-model.cpp" --- "a/examples/talk-llama/llama-model.cpp" +++ "b/examples/talk-llama/llama-model.cpp" @@ -1,128 +1,85 @@ #include "llama-model.h" #include "llama-impl.h" +#include "llama-mmap.h" #include "llama-model-loader.h" -#include "unicode.h" // TODO: remove +#include "ggml-cpp.h" #include #include +#include #include +#include #include #include -static const size_t kiB = 1024; -static const size_t MiB = 1024*kiB; -static const size_t GiB = 1024*MiB; - const char * llm_type_name(llm_type type) { switch (type) { - case MODEL_14M: return "14M"; - case MODEL_17M: return "17M"; - case MODEL_22M: return "22M"; - case MODEL_33M: return "33M"; - case MODEL_60M: return "60M"; - case MODEL_70M: return "70M"; - case MODEL_80M: return "80M"; - case MODEL_109M: return "109M"; - case MODEL_137M: return "137M"; - case MODEL_160M: return "160M"; - case MODEL_220M: return "220M"; - case MODEL_250M: return "250M"; - case MODEL_270M: return "270M"; - case MODEL_335M: return "335M"; - case MODEL_410M: return "410M"; - case MODEL_450M: return "450M"; - case MODEL_770M: return "770M"; - case MODEL_780M: return "780M"; - case MODEL_0_5B: return "0.5B"; - case MODEL_1B: return "1B"; - case MODEL_1_3B: return "1.3B"; - case MODEL_1_4B: return "1.4B"; - case MODEL_1_5B: return "1.5B"; - case MODEL_1_6B: return "1.6B"; - case MODEL_2B: return "2B"; - case MODEL_2_8B: return "2.8B"; - case MODEL_3B: return "3B"; - case MODEL_4B: return "4B"; - case MODEL_6B: return "6B"; - case MODEL_6_9B: return "6.9B"; - case MODEL_7B: return "7B"; - case MODEL_8B: return "8B"; - case MODEL_9B: return "9B"; - case MODEL_11B: return "11B"; - case MODEL_12B: return "12B"; - case MODEL_13B: return "13B"; - case MODEL_14B: return "14B"; - case MODEL_15B: return "15B"; - case MODEL_16B: return "16B"; - case MODEL_20B: return "20B"; - case MODEL_30B: return "30B"; - case MODEL_32B: return "32B"; - case MODEL_34B: return "34B"; - case MODEL_35B: return "35B"; - case MODEL_40B: return "40B"; - case MODEL_65B: return "65B"; - case MODEL_70B: return "70B"; - case MODEL_236B: return "236B"; - case MODEL_314B: return "314B"; - case MODEL_671B: return "671B"; - case MODEL_SMALL: return "0.1B"; - case MODEL_MEDIUM: return "0.4B"; - case MODEL_LARGE: return "0.8B"; - case MODEL_XL: return "1.5B"; - case MODEL_A1_7B: return "A1.7B"; - case MODEL_A2_7B: return "A2.7B"; - case MODEL_8x7B: return "8x7B"; - case MODEL_8x22B: return "8x22B"; - case MODEL_16x12B: return "16x12B"; - case MODEL_10B_128x3_66B: return "10B+128x3.66B"; - case MODEL_57B_A14B: return "57B.A14B"; - case MODEL_27B: return "27B"; - default: return "?B"; - } -} - -static std::string llama_model_ftype_name(llama_ftype ftype) { - if (ftype & LLAMA_FTYPE_GUESSED) { - return llama_model_ftype_name((enum llama_ftype) (ftype & ~LLAMA_FTYPE_GUESSED)) + " (guessed)"; - } - - switch (ftype) { - case LLAMA_FTYPE_ALL_F32: return "all F32"; - case LLAMA_FTYPE_MOSTLY_F16: return "F16"; - case LLAMA_FTYPE_MOSTLY_BF16: return "BF16"; - case LLAMA_FTYPE_MOSTLY_Q4_0: return "Q4_0"; - case LLAMA_FTYPE_MOSTLY_Q4_1: return "Q4_1"; - case LLAMA_FTYPE_MOSTLY_Q5_0: return "Q5_0"; - case LLAMA_FTYPE_MOSTLY_Q5_1: return "Q5_1"; - case LLAMA_FTYPE_MOSTLY_Q8_0: return "Q8_0"; - case LLAMA_FTYPE_MOSTLY_Q2_K: return "Q2_K - Medium"; - case LLAMA_FTYPE_MOSTLY_Q2_K_S: return "Q2_K - Small"; - case LLAMA_FTYPE_MOSTLY_Q3_K_S: return "Q3_K - Small"; - case LLAMA_FTYPE_MOSTLY_Q3_K_M: return "Q3_K - Medium"; - case LLAMA_FTYPE_MOSTLY_Q3_K_L: return "Q3_K - Large"; - case LLAMA_FTYPE_MOSTLY_Q4_K_S: return "Q4_K - Small"; - case LLAMA_FTYPE_MOSTLY_Q4_K_M: return "Q4_K - Medium"; - case LLAMA_FTYPE_MOSTLY_Q5_K_S: return "Q5_K - Small"; - case LLAMA_FTYPE_MOSTLY_Q5_K_M: return "Q5_K - Medium"; - case LLAMA_FTYPE_MOSTLY_Q6_K: return "Q6_K"; - case LLAMA_FTYPE_MOSTLY_TQ1_0: return "TQ1_0 - 1.69 bpw ternary"; - case LLAMA_FTYPE_MOSTLY_TQ2_0: return "TQ2_0 - 2.06 bpw ternary"; - case LLAMA_FTYPE_MOSTLY_IQ2_XXS: return "IQ2_XXS - 2.0625 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ2_XS: return "IQ2_XS - 2.3125 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ2_S: return "IQ2_S - 2.5 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ2_M: return "IQ2_M - 2.7 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ3_XS: return "IQ3_XS - 3.3 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ3_XXS: return "IQ3_XXS - 3.0625 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ1_S: return "IQ1_S - 1.5625 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ1_M: return "IQ1_M - 1.75 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ4_NL: return "IQ4_NL - 4.5 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ4_XS: return "IQ4_XS - 4.25 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ3_S: return "IQ3_S - 3.4375 bpw"; - case LLAMA_FTYPE_MOSTLY_IQ3_M: return "IQ3_S mix - 3.66 bpw"; - - default: return "unknown, may not work"; + case LLM_TYPE_14M: return "14M"; + case LLM_TYPE_17M: return "17M"; + case LLM_TYPE_22M: return "22M"; + case LLM_TYPE_33M: return "33M"; + case LLM_TYPE_60M: return "60M"; + case LLM_TYPE_70M: return "70M"; + case LLM_TYPE_80M: return "80M"; + case LLM_TYPE_109M: return "109M"; + case LLM_TYPE_137M: return "137M"; + case LLM_TYPE_160M: return "160M"; + case LLM_TYPE_220M: return "220M"; + case LLM_TYPE_250M: return "250M"; + case LLM_TYPE_270M: return "270M"; + case LLM_TYPE_335M: return "335M"; + case LLM_TYPE_410M: return "410M"; + case LLM_TYPE_450M: return "450M"; + case LLM_TYPE_770M: return "770M"; + case LLM_TYPE_780M: return "780M"; + case LLM_TYPE_0_5B: return "0.5B"; + case LLM_TYPE_1B: return "1B"; + case LLM_TYPE_1_3B: return "1.3B"; + case LLM_TYPE_1_4B: return "1.4B"; + case LLM_TYPE_1_5B: return "1.5B"; + case LLM_TYPE_1_6B: return "1.6B"; + case LLM_TYPE_2B: return "2B"; + case LLM_TYPE_2_8B: return "2.8B"; + case LLM_TYPE_3B: return "3B"; + case LLM_TYPE_4B: return "4B"; + case LLM_TYPE_6B: return "6B"; + case LLM_TYPE_6_9B: return "6.9B"; + case LLM_TYPE_7B: return "7B"; + case LLM_TYPE_8B: return "8B"; + case LLM_TYPE_9B: return "9B"; + case LLM_TYPE_11B: return "11B"; + case LLM_TYPE_12B: return "12B"; + case LLM_TYPE_13B: return "13B"; + case LLM_TYPE_14B: return "14B"; + case LLM_TYPE_15B: return "15B"; + case LLM_TYPE_16B: return "16B"; + case LLM_TYPE_20B: return "20B"; + case LLM_TYPE_30B: return "30B"; + case LLM_TYPE_32B: return "32B"; + case LLM_TYPE_34B: return "34B"; + case LLM_TYPE_35B: return "35B"; + case LLM_TYPE_40B: return "40B"; + case LLM_TYPE_65B: return "65B"; + case LLM_TYPE_70B: return "70B"; + case LLM_TYPE_236B: return "236B"; + case LLM_TYPE_314B: return "314B"; + case LLM_TYPE_671B: return "671B"; + case LLM_TYPE_SMALL: return "0.1B"; + case LLM_TYPE_MEDIUM: return "0.4B"; + case LLM_TYPE_LARGE: return "0.8B"; + case LLM_TYPE_XL: return "1.5B"; + case LLM_TYPE_A1_7B: return "A1.7B"; + case LLM_TYPE_A2_7B: return "A2.7B"; + case LLM_TYPE_8x7B: return "8x7B"; + case LLM_TYPE_8x22B: return "8x22B"; + case LLM_TYPE_16x12B: return "16x12B"; + case LLM_TYPE_16x3_8B: return "16x3.8B"; + case LLM_TYPE_10B_128x3_66B: return "10B+128x3.66B"; + case LLM_TYPE_57B_A14B: return "57B.A14B"; + case LLM_TYPE_27B: return "27B"; + default: return "?B"; } } @@ -134,132 +91,301 @@ static const char * llama_expert_gating_func_name(llama_expert_gating_func_type } } -std::string llama_model_arch_name (const llama_model & model) { - return llm_arch_name(model.arch); -} +static const std::map LLAMA_ROPE_SCALING_TYPES = { + { LLAMA_ROPE_SCALING_TYPE_NONE, "none" }, + { LLAMA_ROPE_SCALING_TYPE_LINEAR, "linear" }, + { LLAMA_ROPE_SCALING_TYPE_YARN, "yarn" }, + { LLAMA_ROPE_SCALING_TYPE_LONGROPE, "longrope" }, +}; -std::string llama_model_type_name (const llama_model & model) { - return llm_type_name(model.type); -} +static llama_rope_scaling_type llama_rope_scaling_type_from_string(const std::string & name) { + for (const auto & kv : LLAMA_ROPE_SCALING_TYPES) { + if (kv.second == name) { + return (llama_rope_scaling_type) kv.first; + } + } -std::string llama_model_ftype_name(const llama_model & model) { - return llama_model_ftype_name(model.ftype); + return LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED; } -template -static bool buft_supported(ggml_backend_buffer_type_t buft, ggml_backend_dev_t dev, F & fn) { +// checks if the weight tensor can be used with the specified buffer type and device +static bool weight_buft_supported(const llama_hparams & hparams, ggml_tensor * w, ggml_op op, ggml_backend_buffer_type_t buft, ggml_backend_dev_t dev) { + GGML_ASSERT(w != nullptr); + + if (op == GGML_OP_NONE) { + return true; + } + ggml_init_params params = { /*.mem_size =*/ ggml_tensor_overhead()*8, /*.mem_buffer =*/ NULL, /*.no_alloc =*/ true, }; - - ggml_context_ptr ctx { ggml_init(params) }; - if (!ctx) { + ggml_context_ptr ctx_ptr { ggml_init(params) }; + if (!ctx_ptr) { throw std::runtime_error(format("failed to create ggml context")); } + ggml_context * ctx = ctx_ptr.get(); - ggml_backend_buffer_ptr buf { ggml_backend_buft_alloc_buffer(buft, 0) }; - ggml_tensor * op_tensor = fn(ctx.get()); - for (int i = 0; i < GGML_MAX_SRC; i++) { - if (op_tensor->src[i] != nullptr) { - assert(op_tensor->src[i]->buffer == nullptr); - op_tensor->src[i]->buffer = buf.get(); - } + ggml_tensor * op_tensor = nullptr; + + switch (op) { + case GGML_OP_GET_ROWS: + { + ggml_tensor * b = ggml_new_tensor_1d(ctx, GGML_TYPE_I32, 512); + op_tensor = ggml_get_rows(ctx, w, b); + } break; + case GGML_OP_MUL_MAT: + { + ggml_tensor * b = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, w->ne[0], 512, w->ne[2], w->ne[3]); + op_tensor = ggml_mul_mat(ctx, w, b); + } break; + case GGML_OP_MUL_MAT_ID: + { + int n_expert_used = hparams.n_expert_used; + ggml_tensor * b = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, w->ne[0], n_expert_used, 512); + ggml_tensor * ids = ggml_new_tensor_2d(ctx, GGML_TYPE_I32, n_expert_used, 512); + op_tensor = ggml_mul_mat_id(ctx, w, b, ids); + } break; + case GGML_OP_ADD: + { + ggml_tensor * a = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, w->ne[0], w->ne[1], w->ne[2], w->ne[3]); + op_tensor = ggml_add(ctx, a, w); + } break; + case GGML_OP_MUL: + { + ggml_tensor * a = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, w->ne[0], w->ne[1], w->ne[2], w->ne[3]); + op_tensor = ggml_mul(ctx, a, w); + } break; + case GGML_OP_DIV: + { + ggml_tensor * a = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, w->ne[0]); + op_tensor = ggml_div(ctx, a, w); + } break; + case GGML_OP_ROPE: + { + int n_embd_head = hparams.n_embd_head_v; + int n_head = hparams.n_head(); + ggml_tensor * a = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, n_embd_head, n_head, 512); + ggml_tensor * b = ggml_new_tensor_1d(ctx, GGML_TYPE_I32, 512); + op_tensor = ggml_rope_ext( + ctx, a, b, w, + 0, 0, 0, 0, 0, + 0, 0, 0, 0 + ); + + } break; + case GGML_OP_SSM_CONV: + { + // FIXME + ggml_tensor * conv_x = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, 12345, w->ne[1], 6789); + op_tensor = ggml_ssm_conv(ctx, conv_x, w); + } break; + case GGML_OP_SSM_SCAN: + { + // FIXME + const int64_t d_state = w->ne[0]; + const int64_t d_inner = w->ne[1]; + const int64_t n_seq_tokens = 512; + const int64_t n_seqs = 1; + ggml_tensor * s = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, d_state, d_inner, n_seqs); + ggml_tensor * x = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, d_inner, n_seq_tokens, n_seqs); + ggml_tensor * dt = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, d_inner, n_seq_tokens, n_seqs); + ggml_tensor * B = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, d_state, n_seq_tokens, n_seqs); + ggml_tensor * C = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, d_state, n_seq_tokens, n_seqs); + op_tensor = ggml_ssm_scan(ctx, s, x, dt, w, B, C); + } break; + case GGML_OP_RWKV_WKV6: + { + // FIXME + const int64_t S = 123; + const int64_t H = 123; + const int64_t n_tokens = 123; + const int64_t n_seqs = 123; + ggml_tensor * k = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, S, H, n_tokens); + ggml_tensor * v = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, S, H, n_tokens); + ggml_tensor * r = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, S, H, n_tokens); + ggml_tensor * tf = w; + ggml_tensor * td = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, S, H, n_tokens); + ggml_tensor * state = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, S, n_seqs, S, H); + op_tensor = ggml_rwkv_wkv6(ctx, k, v, r, tf, td, state); + } break; + case GGML_OP_IM2COL: + { + const int n_embd = hparams.n_embd; + ggml_tensor * b = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, n_embd, w->ne[1], 1, 1); + op_tensor = ggml_im2col(ctx, w, b, 1, 0, 0, 0, 1, 0, false, GGML_TYPE_F16); + } break; + default: + GGML_ABORT("%s: missing test for op %s for tensor %s", __func__, ggml_op_name(op), w->name); } + // create a temporary dummy buffer for the weight so that supports_op can check the buffer type + GGML_ASSERT(w->buffer == nullptr); + w->buffer = ggml_backend_buft_alloc_buffer(buft, 0); bool op_supported = ggml_backend_dev_supports_op(dev, op_tensor); + ggml_backend_buffer_free(w->buffer); + w->buffer = nullptr; return op_supported; } -template -static ggml_backend_buffer_type_t select_buft(const llama_model::buft_list_t & buft_list, const F & fn) { +// lists of buffer types used for each layer +using buft_list_t = std::vector>; + +// find the first buffer type in the list that can use the tensor +static ggml_backend_buffer_type_t select_weight_buft(const llama_hparams & hparams, ggml_tensor * tensor, ggml_op op, const buft_list_t & buft_list) { + GGML_ASSERT(!buft_list.empty()); for (const auto & cur : buft_list) { ggml_backend_dev_t cur_dev = cur.first; ggml_backend_buffer_type_t cur_buft = cur.second; - if (buft_supported(cur_buft, cur_dev, fn)) { + if (weight_buft_supported(hparams, tensor, op, cur_buft, cur_dev)) { return cur_buft; } } - - throw std::runtime_error(format("no suitable buffer type found")); + return nullptr; } -ggml_backend_buffer_type_t llama_model_select_buft(const llama_model & model, int il) { - return select_buft( - *model.dev_layer.at(il).buft_list, - [&](ggml_context * ctx) { - ggml_tensor * cur = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, model.hparams.n_embd); - ggml_tensor * layer_dir = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, model.hparams.n_embd); - return ggml_add(ctx, cur, layer_dir); - }); -} +// CPU: ACCEL -> CPU extra -> GPU host -> CPU +static buft_list_t make_cpu_buft_list(const std::vector & devices) { + buft_list_t buft_list; + + // add ACCEL buffer types + for (size_t i = 0; i < ggml_backend_dev_count(); ++i) { + ggml_backend_dev_t dev = ggml_backend_dev_get(i); + if (ggml_backend_dev_type(dev) == GGML_BACKEND_DEVICE_TYPE_ACCEL) { + auto * buft = ggml_backend_dev_buffer_type(dev); + // skip + if (buft != ggml_backend_cpu_buffer_type()) { + buft_list.emplace_back(dev, buft); + } + } + } -struct ggml_tensor * llama_model_get_tensor(const struct llama_model & model, const char * name) { - auto it = std::find_if(model.tensors_by_name.begin(), model.tensors_by_name.end(), - [name](const std::pair & it) { - return it.first == name; - }); - if (it == model.tensors_by_name.end()) { - return nullptr; + // add extra buffer types + auto * cpu_dev = ggml_backend_dev_by_type(GGML_BACKEND_DEVICE_TYPE_CPU); + auto * cpu_reg = ggml_backend_dev_backend_reg(cpu_dev); + auto ggml_backend_dev_get_extra_bufts_fn = (ggml_backend_dev_get_extra_bufts_t) + ggml_backend_reg_get_proc_address(cpu_reg, "ggml_backend_dev_get_extra_bufts"); + if (ggml_backend_dev_get_extra_bufts_fn) { + ggml_backend_buffer_type_t * extra_bufts = ggml_backend_dev_get_extra_bufts_fn(cpu_dev); + while (extra_bufts && *extra_bufts) { + buft_list.emplace_back(cpu_dev, *extra_bufts); + ++extra_bufts; + } } - return it->second; -} + // add a host buffer type + // storing the tensors in a host buffer is useful when the processing of large batches + // is offloaded to a GPU device, since it reduces the time spent on data transfers + // generally, this will be done using the first device in the list + // a better approach would be to handle this on a weight-by-weight basis using the offload_op + // function of the device to determine if it would benefit from being stored in a host buffer + for (auto * dev : devices) { + ggml_backend_buffer_type_t buft = ggml_backend_dev_host_buffer_type(dev); + if (buft) { + buft_list.emplace_back(dev, buft); + break; + } + } -size_t llama_model_max_nodes(const llama_model & model) { - return std::max(8192, model.tensors_by_name.size()*5); -} + // add the CPU buffer type + for (size_t i = 0; i < ggml_backend_dev_count(); ++i) { + ggml_backend_dev_t dev = ggml_backend_dev_get(i); + if (ggml_backend_dev_type(dev) == GGML_BACKEND_DEVICE_TYPE_CPU) { + buft_list.emplace_back(dev, ggml_backend_dev_buffer_type(dev)); + } + } -static const std::map LLAMA_ROPE_SCALING_TYPES = { - { LLAMA_ROPE_SCALING_TYPE_NONE, "none" }, - { LLAMA_ROPE_SCALING_TYPE_LINEAR, "linear" }, - { LLAMA_ROPE_SCALING_TYPE_YARN, "yarn" }, - { LLAMA_ROPE_SCALING_TYPE_LONGROPE, "longrope" }, -}; + return buft_list; +} -static llama_rope_scaling_type llama_rope_scaling_type_from_string(const std::string & name) { - for (const auto & kv : LLAMA_ROPE_SCALING_TYPES) { - if (kv.second == name) { - return (llama_rope_scaling_type) kv.first; +// GPU: split if LLAMA_SPLIT_MODE_ROW -> GPU +static buft_list_t make_gpu_buft_list(ggml_backend_dev_t dev, enum llama_split_mode split_mode, const float * tensor_split) { + buft_list_t buft_list; + + // add the device split buffer type if requested and available + if (split_mode == LLAMA_SPLIT_MODE_ROW) { + ggml_backend_reg_t reg = ggml_backend_dev_backend_reg(dev); + auto ggml_backend_split_buffer_type_fn = (ggml_backend_split_buffer_type_t) + ggml_backend_reg_get_proc_address(reg, "ggml_backend_split_buffer_type"); + if (ggml_backend_split_buffer_type_fn) { + size_t dev_index = [&]() { + auto * reg = ggml_backend_dev_backend_reg(dev); + for (size_t i = 0; i < ggml_backend_reg_dev_count(reg); ++i) { + if (ggml_backend_reg_dev_get(reg, i) == dev) { + return i; + } + } + throw std::runtime_error(format("device %s not found in its backend reg", ggml_backend_dev_name(dev))); + }(); + auto * buft = ggml_backend_split_buffer_type_fn(dev_index, tensor_split); + if (buft != nullptr) { + buft_list.emplace_back(dev, buft); + } } } - return LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED; + // add the device default buffer type + buft_list.emplace_back(dev, ggml_backend_dev_buffer_type(dev)); + + return buft_list; } -// NOTE: avoid ever using this except for building the token_to_piece caches -static std::string llama_token_to_piece(const struct llama_model * model, llama_token token, bool special) { - std::string piece; - piece.resize(piece.capacity()); // using string internal cache - const int n_chars = llama_token_to_piece(model, token, &piece[0], piece.size(), 0, special); - if (n_chars < 0) { - piece.resize(-n_chars); - int check = llama_token_to_piece(model, token, &piece[0], piece.size(), 0, special); - GGML_ASSERT(check == -n_chars); - } - else { - piece.resize(n_chars); - } +struct llama_model::impl { + impl() {} + ~impl() {} + + uint64_t n_elements = 0; + + size_t n_bytes = 0; + + std::string desc_str; + + // model memory mapped files + llama_mmaps mappings; + + // objects representing data potentially being locked in memory + llama_mlocks mlock_bufs; + llama_mlocks mlock_mmaps; + + // contexts where the model tensors metadata is stored + std::vector ctxs; + + // the model memory buffers for the tensor data + std::vector bufs; + + buft_list_t cpu_buft_list; + std::map gpu_buft_list; + + struct layer_dev { + ggml_backend_dev_t dev; + buft_list_t * buft_list; + }; - return piece; + layer_dev dev_input = {}; + layer_dev dev_output = {}; + std::vector dev_layer; +}; + +llama_model::llama_model(const struct llama_model_params & params) : params(params), pimpl(std::make_unique()) { } -void llm_load_stats(llama_model_loader & ml, llama_model & model) { - model.n_elements = ml.n_elements; - model.n_bytes = ml.n_bytes; +llama_model::~llama_model() {} + +void llama_model::load_stats(llama_model_loader & ml) { + pimpl->n_elements = ml.n_elements; + pimpl->n_bytes = ml.n_bytes; } -void llm_load_arch(llama_model_loader & ml, llama_model & model) { - model.arch = ml.get_arch(); - if (model.arch == LLM_ARCH_UNKNOWN) { +void llama_model::load_arch(llama_model_loader & ml) { + arch = ml.get_arch(); + if (arch == LLM_ARCH_UNKNOWN) { throw std::runtime_error("unknown model architecture: '" + ml.get_arch_name() + "'"); } } -void llm_load_hparams(llama_model_loader & ml, llama_model & model) { - auto & hparams = model.hparams; +void llama_model::load_hparams(llama_model_loader & ml) { const gguf_context * ctx = ml.meta.get(); // get metadata as string @@ -270,14 +396,11 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { } const char * name = gguf_get_key(ctx, i); const std::string value = gguf_kv_to_str(ctx, i); - model.gguf_kv.emplace(name, value); + gguf_kv.emplace(name, value); } // get general kv - ml.get_key(LLM_KV_GENERAL_NAME, model.name, false); - - // get hparams kv - ml.get_key(LLM_KV_VOCAB_SIZE, hparams.n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, hparams.n_vocab, false); + ml.get_key(LLM_KV_GENERAL_NAME, name, false); // everything past this point is not vocab-related if (hparams.vocab_only) { @@ -290,7 +413,7 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_EXPERT_COUNT, hparams.n_expert, false); ml.get_key(LLM_KV_EXPERT_USED_COUNT, hparams.n_expert_used, false); - if (model.arch == LLM_ARCH_WAVTOKENIZER_DEC) { + if (arch == LLM_ARCH_WAVTOKENIZER_DEC) { ml.get_key(LLM_KV_FEATURES_LENGTH, hparams.n_embd_features); ml.get_key(LLM_KV_POSNET_EMBEDDING_LENGTH, hparams.posnet.n_embd); @@ -363,7 +486,7 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ROPE_DIMENSION_COUNT, hparams.n_rot, false); - if (model.arch == LLM_ARCH_LLAMA || model.arch == LLM_ARCH_DECI || model.arch == LLM_ARCH_FALCON) { + if (arch == LLM_ARCH_LLAMA || arch == LLM_ARCH_DECI || arch == LLM_ARCH_FALCON) { if (hparams.n_rot != hparams.n_embd_head_k) { throw std::runtime_error(format("invalid n_rot: %u, expected %u", hparams.n_rot, hparams.n_embd_head_k)); } @@ -374,34 +497,36 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { hparams.n_embd_head_v = 0; } - using e_model = llm_type; // TMP + // for differentiating model types + uint32_t n_vocab = 0; + ml.get_key(LLM_KV_VOCAB_SIZE, n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, n_vocab, false); // arch-specific KVs - switch (model.arch) { + switch (arch) { case LLM_ARCH_LLAMA: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); if (hparams.n_expert == 8) { switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_8x7B; break; - case 56: model.type = e_model::MODEL_8x22B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_8x7B; break; + case 56: type = LLM_TYPE_8x22B; break; + default: type = LLM_TYPE_UNKNOWN; } } else { switch (hparams.n_layer) { - case 16: model.type = e_model::MODEL_1B; break; // Llama 3.2 1B - case 22: model.type = e_model::MODEL_1B; break; - case 26: model.type = e_model::MODEL_3B; break; - case 28: model.type = e_model::MODEL_3B; break; // Llama 3.2 3B + case 16: type = LLM_TYPE_1B; break; // Llama 3.2 1B + case 22: type = LLM_TYPE_1B; break; + case 26: type = LLM_TYPE_3B; break; + case 28: type = LLM_TYPE_3B; break; // Llama 3.2 3B // granite uses a vocab with len 49152 - case 32: model.type = hparams.n_vocab == 49152 ? e_model::MODEL_3B : (hparams.n_vocab < 40000 ? e_model::MODEL_7B : e_model::MODEL_8B); break; - case 36: model.type = e_model::MODEL_8B; break; // granite - case 40: model.type = e_model::MODEL_13B; break; - case 48: model.type = e_model::MODEL_34B; break; - case 60: model.type = e_model::MODEL_30B; break; - case 80: model.type = hparams.n_head() == hparams.n_head_kv() ? e_model::MODEL_65B : e_model::MODEL_70B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = n_vocab == 49152 ? LLM_TYPE_3B : (n_vocab < 40000 ? LLM_TYPE_7B : LLM_TYPE_8B); break; + case 36: type = LLM_TYPE_8B; break; // granite + case 40: type = LLM_TYPE_13B; break; + case 48: type = LLM_TYPE_34B; break; + case 60: type = LLM_TYPE_30B; break; + case 80: type = hparams.n_head() == hparams.n_head_kv() ? LLM_TYPE_65B : LLM_TYPE_70B; break; + default: type = LLM_TYPE_UNKNOWN; } } } break; @@ -409,33 +534,33 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_7B; break; - case 80: model.type = e_model::MODEL_70B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_7B; break; + case 80: type = LLM_TYPE_70B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_MINICPM: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - ml.get_key(LLM_KV_EMBEDDING_SCALE, hparams.f_embedding_scale); - ml.get_key(LLM_KV_RESIDUAL_SCALE, hparams.f_residual_scale); - ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); + ml.get_key(LLM_KV_EMBEDDING_SCALE, hparams.f_embedding_scale); + ml.get_key(LLM_KV_RESIDUAL_SCALE, hparams.f_residual_scale); + ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); switch (hparams.n_layer) { - case 52: model.type = e_model::MODEL_1B; break; - case 40: model.type = e_model::MODEL_2B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 52: type = LLM_TYPE_1B; break; + case 40: type = LLM_TYPE_2B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_MINICPM3: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - ml.get_key(LLM_KV_ATTENTION_Q_LORA_RANK, hparams.n_lora_q); - ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK, hparams.n_lora_kv); + ml.get_key(LLM_KV_ATTENTION_Q_LORA_RANK, hparams.n_lora_q); + ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK, hparams.n_lora_kv); switch (hparams.n_layer) { - case 62: model.type = e_model::MODEL_4B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 62: type = LLM_TYPE_4B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_GROK: @@ -443,8 +568,8 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 64: model.type = e_model::MODEL_314B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 64: type = LLM_TYPE_314B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_FALCON: @@ -452,21 +577,21 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_7B; break; - case 60: model.type = e_model::MODEL_40B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_7B; break; + case 60: type = LLM_TYPE_40B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_BAICHUAN: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_7B; break; - case 40: model.type = e_model::MODEL_13B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_7B; break; + case 40: type = LLM_TYPE_13B; break; + default: type = LLM_TYPE_UNKNOWN; } - if (model.type == e_model::MODEL_13B) { + if (type == LLM_TYPE_13B) { // TODO: become GGUF KV parameter hparams.f_max_alibi_bias = 8.0f; } @@ -475,19 +600,19 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 24: model.type = e_model::MODEL_1B; break; - case 36: model.type = e_model::MODEL_3B; break; - case 42: model.type = e_model::MODEL_7B; break; - case 40: model.type = e_model::MODEL_15B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 24: type = LLM_TYPE_1B; break; + case 36: type = LLM_TYPE_3B; break; + case 42: type = LLM_TYPE_7B; break; + case 40: type = LLM_TYPE_15B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_REFACT: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_1B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_1B; break; + default: type = LLM_TYPE_UNKNOWN; } // TODO: become GGUF KV parameter @@ -497,48 +622,45 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_ATTENTION_CAUSAL, hparams.causal_attn); - ml.get_key(LLM_KV_TOKENIZER_TOKEN_TYPE_COUNT, hparams.n_vocab_type); ml.get_key(LLM_KV_POOLING_TYPE, hparams.pooling_type, false); switch (hparams.n_layer) { case 3: - model.type = e_model::MODEL_17M; break; // bge-micro + type = LLM_TYPE_17M; break; // bge-micro case 6: - model.type = e_model::MODEL_22M; break; // MiniLM-L6 + type = LLM_TYPE_22M; break; // MiniLM-L6 case 12: switch (hparams.n_embd) { - case 384: model.type = e_model::MODEL_33M; break; // MiniLM-L12, bge-small - case 768: model.type = e_model::MODEL_109M; break; // bge-base - default: model.type = e_model::MODEL_UNKNOWN; + case 384: type = LLM_TYPE_33M; break; // MiniLM-L12, bge-small + case 768: type = LLM_TYPE_109M; break; // bge-base + default: type = LLM_TYPE_UNKNOWN; } break; case 24: - model.type = e_model::MODEL_335M; break; // bge-large - default: model.type = e_model::MODEL_UNKNOWN; + type = LLM_TYPE_335M; break; // bge-large + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_JINA_BERT_V2: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_ATTENTION_CAUSAL, hparams.causal_attn); - ml.get_key(LLM_KV_TOKENIZER_TOKEN_TYPE_COUNT, hparams.n_vocab_type); ml.get_key(LLM_KV_POOLING_TYPE, hparams.pooling_type, false); hparams.f_max_alibi_bias = 8.0f; switch (hparams.n_layer) { - case 4: model.type = e_model::MODEL_33M; break; // jina-embeddings-small - case 12: model.type = e_model::MODEL_137M; break; // jina-embeddings-base - default: model.type = e_model::MODEL_UNKNOWN; + case 4: type = LLM_TYPE_33M; break; // jina-embeddings-small + case 12: type = LLM_TYPE_137M; break; // jina-embeddings-base + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_NOMIC_BERT: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_ATTENTION_CAUSAL, hparams.causal_attn); - ml.get_key(LLM_KV_TOKENIZER_TOKEN_TYPE_COUNT, hparams.n_vocab_type); ml.get_key(LLM_KV_POOLING_TYPE, hparams.pooling_type); if (hparams.n_layer == 12 && hparams.n_embd == 768) { - model.type = e_model::MODEL_137M; + type = LLM_TYPE_137M; } } break; case LLM_ARCH_BLOOM: @@ -546,14 +668,14 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 24: model.type = e_model::MODEL_1B; break; + case 24: type = LLM_TYPE_1B; break; case 30: switch (hparams.n_embd) { - case 2560: model.type = e_model::MODEL_3B; break; - case 4096: model.type = e_model::MODEL_7B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 2560: type = LLM_TYPE_3B; break; + case 4096: type = LLM_TYPE_7B; break; + default: type = LLM_TYPE_UNKNOWN; } break; - default: model.type = e_model::MODEL_UNKNOWN; + default: type = LLM_TYPE_UNKNOWN; } // TODO: become GGUF KV parameter @@ -566,9 +688,9 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_MAX_ALIBI_BIAS, hparams.f_max_alibi_bias); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_7B; break; - case 48: model.type = e_model::MODEL_30B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_7B; break; + case 48: type = LLM_TYPE_30B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_STABLELM: @@ -576,10 +698,10 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 24: model.type = e_model::MODEL_1B; break; - case 32: model.type = e_model::MODEL_3B; break; - case 40: model.type = e_model::MODEL_12B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 24: type = LLM_TYPE_1B; break; + case 32: type = LLM_TYPE_3B; break; + case 40: type = LLM_TYPE_12B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_QWEN: @@ -587,9 +709,9 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_7B; break; - case 40: model.type = e_model::MODEL_13B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_7B; break; + case 40: type = LLM_TYPE_13B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_QWEN2VL: @@ -601,27 +723,27 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 24: model.type = hparams.n_embd == 1024 ? e_model::MODEL_0_5B : e_model::MODEL_1B; break; - case 28: model.type = hparams.n_embd == 1536 ? e_model::MODEL_1_5B : e_model::MODEL_7B; break; - case 32: model.type = e_model::MODEL_7B; break; - case 36: model.type = e_model::MODEL_3B; break; - case 40: model.type = hparams.n_head() == 20 ? e_model::MODEL_4B : e_model::MODEL_13B; break; - case 48: model.type = e_model::MODEL_14B; break; - case 64: model.type = e_model::MODEL_32B; break; - case 80: model.type = e_model::MODEL_70B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 24: type = hparams.n_embd == 1024 ? LLM_TYPE_0_5B : LLM_TYPE_1B; break; + case 28: type = hparams.n_embd == 1536 ? LLM_TYPE_1_5B : LLM_TYPE_7B; break; + case 32: type = LLM_TYPE_7B; break; + case 36: type = LLM_TYPE_3B; break; + case 40: type = hparams.n_head() == 20 ? LLM_TYPE_4B : LLM_TYPE_13B; break; + case 48: type = LLM_TYPE_14B; break; + case 64: type = LLM_TYPE_32B; break; + case 80: type = LLM_TYPE_70B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_QWEN2MOE: { - ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false); + ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp, false); ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 24: model.type = e_model::MODEL_A2_7B; break; - case 28: model.type = e_model::MODEL_57B_A14B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 24: type = LLM_TYPE_A2_7B; break; + case 28: type = LLM_TYPE_57B_A14B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_PHI2: @@ -629,9 +751,9 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 24: model.type = e_model::MODEL_1B; break; - case 32: model.type = e_model::MODEL_3B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 24: type = LLM_TYPE_1B; break; + case 32: type = LLM_TYPE_3B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_PHI3: @@ -639,10 +761,10 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 24: model.type = e_model::MODEL_1B; break; - case 32: model.type = e_model::MODEL_3B; break; - case 40: model.type = e_model::MODEL_14B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 24: type = LLM_TYPE_1B; break; + case 32: type = LLM_TYPE_3B; break; + case 40: type = LLM_TYPE_14B; break; + default: type = LLM_TYPE_UNKNOWN; } // for backward compatibility ; see: https://github.com/ggerganov/llama.cpp/pull/8931 @@ -661,32 +783,41 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { throw std::runtime_error("invalid value for sliding_window"); } } break; + case LLM_ARCH_PHIMOE: + { + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + + switch (hparams.n_layer) { + case 32: type = LLM_TYPE_16x3_8B; break; + default: type = LLM_TYPE_UNKNOWN; + } + } break; case LLM_ARCH_PLAMO: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 40: model.type = e_model::MODEL_13B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 40: type = LLM_TYPE_13B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_GPT2: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 12: model.type = e_model::MODEL_SMALL; break; - case 24: model.type = e_model::MODEL_MEDIUM; break; - case 36: model.type = e_model::MODEL_LARGE; break; - case 48: model.type = e_model::MODEL_XL; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 12: type = LLM_TYPE_SMALL; break; + case 24: type = LLM_TYPE_MEDIUM; break; + case 36: type = LLM_TYPE_LARGE; break; + case 48: type = LLM_TYPE_XL; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_CODESHELL: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 42: model.type = e_model::MODEL_7B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 42: type = LLM_TYPE_7B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_ORION: @@ -694,17 +825,17 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 40: model.type = e_model::MODEL_14B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 40: type = LLM_TYPE_14B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_INTERNLM2: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_7B; break; - case 48: model.type = e_model::MODEL_20B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_7B; break; + case 48: type = LLM_TYPE_20B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_GEMMA: @@ -712,37 +843,37 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 18: model.type = e_model::MODEL_2B; break; - case 28: model.type = e_model::MODEL_7B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 18: type = LLM_TYPE_2B; break; + case 28: type = LLM_TYPE_7B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_GEMMA2: { hparams.n_swa = 4096; // default value of gemma 2 - ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false); + ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - ml.get_key(LLM_KV_ATTN_LOGIT_SOFTCAPPING, hparams.f_attn_logit_softcapping, false); - ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING, hparams.f_final_logit_softcapping, false); + ml.get_key(LLM_KV_ATTN_LOGIT_SOFTCAPPING, hparams.f_attn_logit_softcapping, false); + ml.get_key(LLM_KV_FINAL_LOGIT_SOFTCAPPING, hparams.f_final_logit_softcapping, false); hparams.attn_soft_cap = true; switch (hparams.n_layer) { - case 26: model.type = e_model::MODEL_2B; break; - case 42: model.type = e_model::MODEL_9B; break; - case 46: model.type = e_model::MODEL_27B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 26: type = LLM_TYPE_2B; break; + case 42: type = LLM_TYPE_9B; break; + case 46: type = LLM_TYPE_27B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_STARCODER2: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 30: model.type = e_model::MODEL_3B; break; - case 32: model.type = e_model::MODEL_7B; break; - case 40: model.type = e_model::MODEL_15B; break; - case 52: model.type = e_model::MODEL_20B; break; // granite - case 88: model.type = e_model::MODEL_34B; break; // granite - default: model.type = e_model::MODEL_UNKNOWN; + case 30: type = LLM_TYPE_3B; break; + case 32: type = LLM_TYPE_7B; break; + case 40: type = LLM_TYPE_15B; break; + case 52: type = LLM_TYPE_20B; break; // granite + case 88: type = LLM_TYPE_34B; break; // granite + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_MAMBA: @@ -758,51 +889,51 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { switch (hparams.n_layer) { case 24: switch (hparams.n_embd) { - case 768: model.type = e_model::MODEL_SMALL; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 768: type = LLM_TYPE_SMALL; break; + default: type = LLM_TYPE_UNKNOWN; } break; case 48: switch (hparams.n_embd) { - case 1024: model.type = e_model::MODEL_MEDIUM; break; - case 1536: model.type = e_model::MODEL_LARGE; break; - case 2048: model.type = e_model::MODEL_XL; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 1024: type = LLM_TYPE_MEDIUM; break; + case 1536: type = LLM_TYPE_LARGE; break; + case 2048: type = LLM_TYPE_XL; break; + default: type = LLM_TYPE_UNKNOWN; } break; case 64: switch (hparams.n_embd) { - case 2560: model.type = e_model::MODEL_3B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 2560: type = LLM_TYPE_3B; break; + default: type = LLM_TYPE_UNKNOWN; } break; - default: model.type = e_model::MODEL_UNKNOWN; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_XVERSE: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_7B; break; - case 40: model.type = e_model::MODEL_13B; break; - case 80: model.type = e_model::MODEL_65B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_7B; break; + case 40: type = LLM_TYPE_13B; break; + case 80: type = LLM_TYPE_65B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_COMMAND_R: { - ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); + ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 40: model.type = e_model::MODEL_35B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 40: type = LLM_TYPE_35B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_COHERE2: { ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa); - ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); - ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); + ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_8B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_8B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_DBRX: @@ -811,8 +942,8 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv); switch (hparams.n_layer) { - case 40: model.type = e_model::MODEL_16x12B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 40: type = LLM_TYPE_16x12B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_OLMO: @@ -821,10 +952,10 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv, false); switch (hparams.n_layer) { - case 22: model.type = e_model::MODEL_1B; break; - case 32: model.type = e_model::MODEL_7B; break; - case 80: model.type = e_model::MODEL_70B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 22: type = LLM_TYPE_1B; break; + case 32: type = LLM_TYPE_7B; break; + case 80: type = LLM_TYPE_70B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_OLMO2: @@ -832,18 +963,18 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 16: model.type = e_model::MODEL_1B; break; - case 32: model.type = e_model::MODEL_7B; break; - case 40: model.type = e_model::MODEL_13B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 16: type = LLM_TYPE_1B; break; + case 32: type = LLM_TYPE_7B; break; + case 40: type = LLM_TYPE_13B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_OLMOE: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 16: model.type = e_model::MODEL_A1_7B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 16: type = LLM_TYPE_A1_7B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_OPENELM: @@ -851,57 +982,57 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 16: model.type = e_model::MODEL_270M; break; - case 20: model.type = e_model::MODEL_450M; break; - case 28: model.type = e_model::MODEL_1B; break; - case 36: model.type = e_model::MODEL_3B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 16: type = LLM_TYPE_270M; break; + case 20: type = LLM_TYPE_450M; break; + case 28: type = LLM_TYPE_1B; break; + case 36: type = LLM_TYPE_3B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_GPTNEOX: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - ml.get_key(LLM_KV_USE_PARALLEL_RESIDUAL, hparams.use_par_res); + ml.get_key(LLM_KV_USE_PARALLEL_RESIDUAL, hparams.use_par_res); switch (hparams.n_layer) { case 6: switch (hparams.n_ff()) { - case 512: model.type = e_model::MODEL_14M; break; - case 2048: model.type = e_model::MODEL_70M; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 512: type = LLM_TYPE_14M; break; + case 2048: type = LLM_TYPE_70M; break; + default: type = LLM_TYPE_UNKNOWN; } break; case 12: switch (hparams.n_ff()) { - case 3072: model.type = e_model::MODEL_160M; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 3072: type = LLM_TYPE_160M; break; + default: type = LLM_TYPE_UNKNOWN; } break; case 16: switch (hparams.n_ff()) { - case 8192: model.type = e_model::MODEL_1B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 8192: type = LLM_TYPE_1B; break; + default: type = LLM_TYPE_UNKNOWN; } break; case 24: switch (hparams.n_ff()) { - case 4096: model.type = e_model::MODEL_410M; break; - case 8192: model.type = e_model::MODEL_1_4B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 4096: type = LLM_TYPE_410M; break; + case 8192: type = LLM_TYPE_1_4B; break; + default: type = LLM_TYPE_UNKNOWN; } break; case 32: switch (hparams.n_ff()) { - case 10240: model.type = e_model::MODEL_2_8B; break; - case 16384: model.type = e_model::MODEL_6_9B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 10240: type = LLM_TYPE_2_8B; break; + case 16384: type = LLM_TYPE_6_9B; break; + default: type = LLM_TYPE_UNKNOWN; } break; case 36: switch (hparams.n_ff()) { - case 20480: model.type = e_model::MODEL_12B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 20480: type = LLM_TYPE_12B; break; + default: type = LLM_TYPE_UNKNOWN; } break; case 44: switch (hparams.n_ff()) { - case 24576: model.type = e_model::MODEL_20B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 24576: type = LLM_TYPE_20B; break; + default: type = LLM_TYPE_UNKNOWN; } break; - default: model.type = e_model::MODEL_UNKNOWN; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_ARCTIC: @@ -910,40 +1041,40 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { if (hparams.n_expert == 128) { switch (hparams.n_layer) { - case 35: model.type = e_model::MODEL_10B_128x3_66B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 35: type = LLM_TYPE_10B_128x3_66B; break; + default: type = LLM_TYPE_UNKNOWN; } } else { - model.type = e_model::MODEL_UNKNOWN; + type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_DEEPSEEK: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead); - ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp); - ml.get_key(LLM_KV_EXPERT_SHARED_COUNT, hparams.n_expert_shared); - ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale); + ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead); + ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp); + ml.get_key(LLM_KV_EXPERT_SHARED_COUNT, hparams.n_expert_shared); + ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale); switch (hparams.n_layer) { - case 28: model.type = e_model::MODEL_20B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 28: type = LLM_TYPE_20B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_DEEPSEEK2: { bool is_lite = (hparams.n_layer == 27); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead); + ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead); if (!is_lite) { ml.get_key(LLM_KV_ATTENTION_Q_LORA_RANK, hparams.n_lora_q); } - ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK, hparams.n_lora_kv); + ml.get_key(LLM_KV_ATTENTION_KV_LORA_RANK, hparams.n_lora_kv); ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp); - ml.get_key(LLM_KV_EXPERT_SHARED_COUNT, hparams.n_expert_shared); - ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale); - ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM, hparams.expert_weights_norm, false); - ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func, false); + ml.get_key(LLM_KV_EXPERT_SHARED_COUNT, hparams.n_expert_shared); + ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale); + ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM, hparams.expert_weights_norm, false); + ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func, false); if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) { // for compatibility with existing DeepSeek V2 and V2.5 GGUFs // that have no expert_gating_func model parameter set @@ -952,19 +1083,19 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul); switch (hparams.n_layer) { - case 27: model.type = e_model::MODEL_16B; break; - case 60: model.type = e_model::MODEL_236B; break; - case 61: model.type = e_model::MODEL_671B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 27: type = LLM_TYPE_16B; break; + case 60: type = LLM_TYPE_236B; break; + case 61: type = LLM_TYPE_671B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_CHATGLM: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 28: model.type = e_model::MODEL_6B; break; - case 40: model.type = e_model::MODEL_9B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 28: type = LLM_TYPE_6B; break; + case 40: type = LLM_TYPE_9B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_BITNET: @@ -972,13 +1103,13 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 26: model.type = e_model::MODEL_3B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 26: type = LLM_TYPE_3B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_T5: { - ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); ml.get_key(LLM_KV_ATTENTION_RELATIVE_BUCKETS_COUNT, hparams.n_rel_attn_bkts); uint32_t dec_start_token_id; @@ -987,32 +1118,32 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { } switch (hparams.n_layer) { - case 6: model.type = e_model::MODEL_60M; break; // t5-small - case 8: model.type = e_model::MODEL_80M; break; // flan-t5-small + case 6: type = LLM_TYPE_60M; break; // t5-small + case 8: type = LLM_TYPE_80M; break; // flan-t5-small case 12: switch (hparams.n_ff()) { - case 3072: model.type = e_model::MODEL_220M; break; // t5-base - case 2048: model.type = e_model::MODEL_250M; break; // flan-t5-base - default: model.type = e_model::MODEL_UNKNOWN; + case 3072: type = LLM_TYPE_220M; break; // t5-base + case 2048: type = LLM_TYPE_250M; break; // flan-t5-base + default: type = LLM_TYPE_UNKNOWN; } break; case 24: switch (hparams.n_ff()) { - case 4096: model.type = e_model::MODEL_770M; break; // t5-large - case 2816: model.type = e_model::MODEL_780M; break; // flan-t5-large - case 16384: model.type = e_model::MODEL_3B; break; // t5-3b - case 5120: model.type = e_model::MODEL_3B; break; // flan-t5-xl - case 65536: model.type = e_model::MODEL_11B; break; // t5-11b - case 10240: model.type = e_model::MODEL_11B; break; // flan-t5-xxl - default: model.type = e_model::MODEL_UNKNOWN; + case 4096: type = LLM_TYPE_770M; break; // t5-large + case 2816: type = LLM_TYPE_780M; break; // flan-t5-large + case 16384: type = LLM_TYPE_3B; break; // t5-3b + case 5120: type = LLM_TYPE_3B; break; // flan-t5-xl + case 65536: type = LLM_TYPE_11B; break; // t5-11b + case 10240: type = LLM_TYPE_11B; break; // flan-t5-xxl + default: type = LLM_TYPE_UNKNOWN; } break; - default: model.type = e_model::MODEL_UNKNOWN; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_T5ENCODER: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); ml.get_key(LLM_KV_ATTENTION_RELATIVE_BUCKETS_COUNT, hparams.n_rel_attn_bkts); - model.type = e_model::MODEL_UNKNOWN; + type = LLM_TYPE_UNKNOWN; } break; case LLM_ARCH_JAIS: { @@ -1020,18 +1151,18 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_MAX_ALIBI_BIAS, hparams.f_max_alibi_bias); switch (hparams.n_layer) { - case 24: model.type = e_model::MODEL_1_3B; break; - case 40: model.type = e_model::MODEL_13B; break; + case 24: type = LLM_TYPE_1_3B; break; + case 40: type = LLM_TYPE_13B; break; /* TODO: add variants */ - default: model.type = e_model::MODEL_UNKNOWN; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_NEMOTRON: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_4B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_4B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_EXAONE: @@ -1039,44 +1170,48 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_8B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_8B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_RWKV6: + case LLM_ARCH_RWKV6QWEN2: { - ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); - ml.get_key(LLM_KV_WKV_HEAD_SIZE, hparams.wkv_head_size); - ml.get_key(LLM_KV_TIME_MIX_EXTRA_DIM, hparams.time_mix_extra_dim); - ml.get_key(LLM_KV_TIME_DECAY_EXTRA_DIM, hparams.time_decay_extra_dim); - ml.get_key(LLM_KV_RESCALE_EVERY_N_LAYERS, hparams.rescale_every_n_layers, false); + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps, false); + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps, false); + ml.get_key(LLM_KV_WKV_HEAD_SIZE, hparams.wkv_head_size); + ml.get_key(LLM_KV_TIME_MIX_EXTRA_DIM, hparams.time_mix_extra_dim); + ml.get_key(LLM_KV_TIME_DECAY_EXTRA_DIM, hparams.time_decay_extra_dim); + ml.get_key(LLM_KV_RESCALE_EVERY_N_LAYERS, hparams.rescale_every_n_layers, false); + ml.get_key(LLM_KV_TOKEN_SHIFT_COUNT, hparams.token_shift_count, false); switch (hparams.n_layer) { - case 24: model.type = e_model::MODEL_1_6B; break; + case 24: type = LLM_TYPE_1_6B; break; case 32: switch (hparams.n_embd) { - case 2560: model.type = e_model::MODEL_3B; break; - case 4096: model.type = e_model::MODEL_7B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 2560: type = LLM_TYPE_3B; break; + case 4096: type = LLM_TYPE_7B; break; + default: type = LLM_TYPE_UNKNOWN; } break; - case 61: model.type = e_model::MODEL_14B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 61: type = LLM_TYPE_14B; break; + case 64: type = LLM_TYPE_32B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_GRANITE: case LLM_ARCH_GRANITE_MOE: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); - ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); - ml.get_key(LLM_KV_RESIDUAL_SCALE, hparams.f_residual_scale); - ml.get_key(LLM_KV_EMBEDDING_SCALE, hparams.f_embedding_scale); - ml.get_key(LLM_KV_ATTENTION_SCALE, hparams.f_attention_scale); + ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale); + ml.get_key(LLM_KV_RESIDUAL_SCALE, hparams.f_residual_scale); + ml.get_key(LLM_KV_EMBEDDING_SCALE, hparams.f_embedding_scale); + ml.get_key(LLM_KV_ATTENTION_SCALE, hparams.f_attention_scale); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_3B; break; - case 40: model.type = e_model::MODEL_3B; break; + case 32: type = LLM_TYPE_3B; break; + case 40: type = LLM_TYPE_3B; break; // Add additional layer/vocab/etc checks here for other model sizes - default: model.type = e_model::MODEL_UNKNOWN; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_CHAMELEON: @@ -1086,9 +1221,9 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { ml.get_key(LLM_KV_SWIN_NORM, hparams.swin_norm); switch (hparams.n_layer) { - case 32: model.type = e_model::MODEL_7B; break; - case 48: model.type = e_model::MODEL_34B; break; - default: model.type = e_model::MODEL_UNKNOWN; + case 32: type = LLM_TYPE_7B; break; + case 48: type = LLM_TYPE_34B; break; + default: type = LLM_TYPE_UNKNOWN; } } break; case LLM_ARCH_WAVTOKENIZER_DEC: @@ -1101,753 +1236,2309 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) { default: throw std::runtime_error("unsupported model architecture"); } - model.ftype = ml.ftype; + pimpl->n_bytes = ml.n_bytes; + + pimpl->desc_str = arch_name() + " " + type_name() + " " + ml.ftype_name(); if (hparams.f_max_alibi_bias > 0.0f) { hparams.use_alibi = true; } - hparams.rope_type = llama_rope_type(&model); + hparams.rope_type = llama_model_rope_type(this); } -void llm_load_vocab(llama_model_loader & ml, llama_model & model) { - auto & vocab = model.vocab; +void llama_model::load_vocab(llama_model_loader & ml) { + const auto kv = LLM_KV(arch); - struct gguf_context * ctx = ml.meta.get(); - - const auto kv = LLM_KV(model.arch); - - // determine vocab type - { - std::string tokenizer_model; - std::string tokenizer_pre; - - ml.get_key(LLM_KV_TOKENIZER_MODEL, tokenizer_model); - ml.get_key(LLM_KV_TOKENIZER_PRE, tokenizer_pre, false); - - if (tokenizer_model == "no_vocab" || tokenizer_model == "none") { - vocab.type = LLAMA_VOCAB_TYPE_NONE; - - // default special tokens - vocab.special_bos_id = LLAMA_TOKEN_NULL; - vocab.special_eos_id = LLAMA_TOKEN_NULL; - vocab.special_unk_id = LLAMA_TOKEN_NULL; - vocab.special_sep_id = LLAMA_TOKEN_NULL; - vocab.special_pad_id = LLAMA_TOKEN_NULL; - vocab.special_cls_id = LLAMA_TOKEN_NULL; - vocab.special_mask_id = LLAMA_TOKEN_NULL; - vocab.linefeed_id = LLAMA_TOKEN_NULL; - - // read vocab size from metadata - if (!ml.get_key(LLM_KV_VOCAB_SIZE, vocab.n_vocab, false)) { - vocab.n_vocab = 0; - LLAMA_LOG_WARN("%s: there is no vocab_size in metadata, vocab.n_vocab will be set to %u\n", __func__, vocab.n_vocab); - } - return; - } - - if (tokenizer_model == "llama") { - vocab.type = LLAMA_VOCAB_TYPE_SPM; - - // default special tokens - vocab.special_bos_id = 1; - vocab.special_eos_id = 2; - vocab.special_unk_id = 0; - vocab.special_sep_id = LLAMA_TOKEN_NULL; - vocab.special_pad_id = LLAMA_TOKEN_NULL; - vocab.special_cls_id = LLAMA_TOKEN_NULL; - vocab.special_mask_id = LLAMA_TOKEN_NULL; - } else if (tokenizer_model == "bert") { - vocab.type = LLAMA_VOCAB_TYPE_WPM; - - // default special tokens - vocab.special_bos_id = LLAMA_TOKEN_NULL; - vocab.special_eos_id = LLAMA_TOKEN_NULL; - vocab.special_unk_id = 100; - vocab.special_sep_id = 102; - vocab.special_pad_id = 0; - vocab.special_cls_id = 101; - vocab.special_mask_id = 103; - } else if (tokenizer_model == "gpt2") { - vocab.type = LLAMA_VOCAB_TYPE_BPE; - - // read bpe merges and populate bpe ranks - const int merges_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_MERGES).c_str()); - if (merges_keyidx == -1) { - throw std::runtime_error("cannot find tokenizer merges in model file\n"); - } - - const int n_merges = gguf_get_arr_n(ctx, merges_keyidx); - for (int i = 0; i < n_merges; i++) { - const std::string word = gguf_get_arr_str(ctx, merges_keyidx, i); - GGML_ASSERT(unicode_cpts_from_utf8(word).size() > 0); - - std::string first; - std::string second; - - const size_t pos = word.find(' ', 1); - - if (pos != std::string::npos) { - first = word.substr(0, pos); - second = word.substr(pos + 1); - } + vocab.load(ml, kv); +} - vocab.bpe_ranks.emplace(std::make_pair(first, second), i); - } +bool llama_model::load_tensors(llama_model_loader & ml) { + const auto & split_mode = params.split_mode; + const auto & n_gpu_layers = params.n_gpu_layers; + const auto & use_mlock = params.use_mlock; + const auto & tensor_split = params.tensor_split; - // default special tokens - vocab.special_bos_id = 11; - vocab.special_eos_id = 11; - vocab.special_unk_id = LLAMA_TOKEN_NULL; - vocab.special_sep_id = LLAMA_TOKEN_NULL; - vocab.special_pad_id = LLAMA_TOKEN_NULL; - vocab.special_cls_id = LLAMA_TOKEN_NULL; - vocab.special_mask_id = LLAMA_TOKEN_NULL; - } else if (tokenizer_model == "t5") { - vocab.type = LLAMA_VOCAB_TYPE_UGM; - - // default special tokens - vocab.special_bos_id = LLAMA_TOKEN_NULL; - vocab.special_eos_id = 1; - vocab.special_unk_id = 2; - vocab.special_sep_id = LLAMA_TOKEN_NULL; - vocab.special_pad_id = 0; - vocab.special_cls_id = LLAMA_TOKEN_NULL; - vocab.special_mask_id = LLAMA_TOKEN_NULL; - - const int precompiled_charsmap_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_PRECOMPILED_CHARSMAP).c_str()); - if (precompiled_charsmap_keyidx != -1) { - size_t n_precompiled_charsmap = gguf_get_arr_n(ctx, precompiled_charsmap_keyidx); - const char * precompiled_charsmap = (const char *) gguf_get_arr_data(ctx, precompiled_charsmap_keyidx); - vocab.precompiled_charsmap.assign(precompiled_charsmap, precompiled_charsmap + n_precompiled_charsmap); -#ifdef IS_BIG_ENDIAN - // correct endiannes of data in precompiled_charsmap binary blob - uint32_t * xcda_blob_size = (uint32_t *) &vocab.precompiled_charsmap[0]; - *xcda_blob_size = __builtin_bswap32(*xcda_blob_size); - assert(*xcda_blob_size + sizeof(uint32_t) < n_precompiled_charsmap); - size_t xcda_array_size = *xcda_blob_size / sizeof(uint32_t); - uint32_t * xcda_array = (uint32_t *) &vocab.precompiled_charsmap[sizeof(uint32_t)]; - for (size_t i = 0; i < xcda_array_size; ++i) { - xcda_array[i] = __builtin_bswap32(xcda_array[i]); - } -#endif - } - } else if (tokenizer_model == "rwkv") { - vocab.type = LLAMA_VOCAB_TYPE_RWKV; - - // default special tokens - vocab.special_bos_id = LLAMA_TOKEN_NULL; - vocab.special_eos_id = LLAMA_TOKEN_NULL; - vocab.special_unk_id = LLAMA_TOKEN_NULL; - vocab.special_sep_id = LLAMA_TOKEN_NULL; - vocab.special_pad_id = LLAMA_TOKEN_NULL; - } else { - throw std::runtime_error(format("unknown tokenizer: '%s'", tokenizer_model.c_str())); - } + const int n_layer = hparams.n_layer; - // for now, only BPE models have pre-tokenizers - if (vocab.type == LLAMA_VOCAB_TYPE_BPE) { - vocab.tokenizer_add_space_prefix = false; - vocab.tokenizer_clean_spaces = true; - if (tokenizer_pre.empty()) { - LLAMA_LOG_WARN("%s: missing pre-tokenizer type, using: 'default'\n", __func__); - LLAMA_LOG_WARN("%s: \n", __func__); - LLAMA_LOG_WARN("%s: ************************************ \n", __func__); - LLAMA_LOG_WARN("%s: GENERATION QUALITY WILL BE DEGRADED! \n", __func__); - LLAMA_LOG_WARN("%s: CONSIDER REGENERATING THE MODEL \n", __func__); - LLAMA_LOG_WARN("%s: ************************************ \n", __func__); - LLAMA_LOG_WARN("%s: \n", __func__); - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; - } else if (tokenizer_pre == "default") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; - } else if ( - tokenizer_pre == "llama3" || - tokenizer_pre == "llama-v3" || - tokenizer_pre == "llama-bpe"|| - tokenizer_pre == "falcon3") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_LLAMA3; - vocab.tokenizer_ignore_merges = true; - vocab.tokenizer_add_bos = true; - } else if ( - tokenizer_pre == "deepseek-llm") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_LLM; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "deepseek-coder") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_CODER; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "deepseek-v3") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEEPSEEK3_LLM; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "falcon") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_FALCON; - } else if ( - tokenizer_pre == "mpt") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_MPT; - } else if ( - tokenizer_pre == "starcoder") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_STARCODER; - } else if ( - tokenizer_pre == "gpt-2" || - tokenizer_pre == "phi-2" || - tokenizer_pre == "jina-es" || - tokenizer_pre == "jina-de" || - tokenizer_pre == "gigachat" || - tokenizer_pre == "jina-v1-en" || - tokenizer_pre == "jina-v2-es" || - tokenizer_pre == "jina-v2-de" || - tokenizer_pre == "jina-v2-code" || - tokenizer_pre == "roberta-bpe") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_GPT2; - } else if ( - tokenizer_pre == "refact") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_REFACT; - } else if ( - tokenizer_pre == "command-r") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_COMMAND_R; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "qwen2") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_QWEN2; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "stablelm2") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_STABLELM2; - } else if ( - tokenizer_pre == "olmo") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_OLMO; - } else if ( - tokenizer_pre == "dbrx") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DBRX; - } else if ( - tokenizer_pre == "smaug-bpe") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_SMAUG; - } else if ( - tokenizer_pre == "poro-chat") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_PORO; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "chatglm-bpe") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_CHATGLM4; - vocab.special_bos_id = LLAMA_TOKEN_NULL; - } else if ( - tokenizer_pre == "viking") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_VIKING; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "jais") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_JAIS; - } else if ( - tokenizer_pre == "tekken") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_TEKKEN; - vocab.tokenizer_clean_spaces = false; - vocab.tokenizer_ignore_merges = true; - vocab.tokenizer_add_bos = true; - } else if ( - tokenizer_pre == "smollm") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_SMOLLM; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "codeshell") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_CODESHELL; - } else if ( - tokenizer_pre == "bloom") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_BLOOM; - } else if ( - tokenizer_pre == "gpt3-finnish") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH; - } else if ( - tokenizer_pre == "exaone") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_EXAONE; - } else if ( - tokenizer_pre == "chameleon") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_CHAMELEON; - vocab.tokenizer_add_bos = true; - vocab.tokenizer_clean_spaces = false; - } else if ( - tokenizer_pre == "minerva-7b") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_MINERVA; - } else if ( - tokenizer_pre == "megrez") { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_QWEN2; - } else { - throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str())); - } - } else if (vocab.type == LLAMA_VOCAB_TYPE_SPM) { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; - vocab.tokenizer_add_space_prefix = true; - vocab.tokenizer_clean_spaces = false; - vocab.tokenizer_add_bos = true; - vocab.tokenizer_add_eos = false; - } else if (vocab.type == LLAMA_VOCAB_TYPE_WPM) { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; - vocab.tokenizer_add_space_prefix = false; - vocab.tokenizer_clean_spaces = true; - vocab.tokenizer_add_bos = true; - vocab.tokenizer_add_eos = false; - } else if (vocab.type == LLAMA_VOCAB_TYPE_UGM) { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; - vocab.tokenizer_add_bos = false; - vocab.tokenizer_add_eos = true; - } else if (vocab.type == LLAMA_VOCAB_TYPE_RWKV) { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; - vocab.tokenizer_add_space_prefix = false; - vocab.tokenizer_clean_spaces = false; - vocab.tokenizer_add_bos = false; - vocab.tokenizer_add_eos = false; - } else { - vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; - } + const bool use_mmap_buffer = true; - ml.get_key(LLM_KV_TOKENIZER_ADD_PREFIX, vocab.tokenizer_add_space_prefix, false); - ml.get_key(LLM_KV_TOKENIZER_REMOVE_EXTRA_WS, vocab.tokenizer_remove_extra_whitespaces, false); + // build a list of buffer types for the CPU and GPU devices + pimpl->cpu_buft_list = make_cpu_buft_list(devices); + for (auto * dev : devices) { + buft_list_t buft_list = make_gpu_buft_list(dev, split_mode, tensor_split); + // add CPU buffer types as a fallback + buft_list.insert(buft_list.end(), pimpl->cpu_buft_list.begin(), pimpl->cpu_buft_list.end()); + pimpl->gpu_buft_list.emplace(dev, std::move(buft_list)); } - const int token_idx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_LIST).c_str()); - if (token_idx == -1) { - throw std::runtime_error("cannot find tokenizer vocab in model file\n"); + // calculate the split points + bool all_zero = tensor_split == nullptr || std::all_of(tensor_split, tensor_split + n_devices(), [](float x) { return x == 0.0f; }); + std::vector splits(n_devices()); + if (all_zero) { + // default split, by free memory + for (size_t i = 0; i < n_devices(); ++i) { + ggml_backend_dev_t dev = devices[i]; + size_t total; + size_t free; + ggml_backend_dev_memory(dev, &free, &total); + splits[i] = free; + } + } else { + std::copy(tensor_split, tensor_split + n_devices(), splits.begin()); } - const float * scores = nullptr; - const int score_idx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_SCORES).c_str()); - if (score_idx != -1) { - scores = (const float * ) gguf_get_arr_data(ctx, score_idx); + // sum and normalize the splits to get the split points + float split_sum = 0.0f; + for (size_t i = 0; i < n_devices(); ++i) { + split_sum += splits[i]; + splits[i] = split_sum; } - - const int * toktypes = nullptr; - const int toktype_idx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_TOKEN_TYPE).c_str()); - if (toktype_idx != -1) { - toktypes = (const int * ) gguf_get_arr_data(ctx, toktype_idx); + for (size_t i = 0; i < n_devices(); ++i) { + splits[i] /= split_sum; } - const uint32_t n_vocab = gguf_get_arr_n(ctx, token_idx); + ggml_backend_dev_t cpu_dev = ggml_backend_dev_by_type(GGML_BACKEND_DEVICE_TYPE_CPU); + const int i_gpu_start = std::max((int) hparams.n_layer - n_gpu_layers, (int) 0); + const int act_gpu_layers = devices.empty() ? 0 : std::min(n_gpu_layers, (int)n_layer + 1); + auto get_layer_buft_list = [&](int il) -> llama_model::impl::layer_dev { + if (il < i_gpu_start || (il - i_gpu_start) >= act_gpu_layers) { + return {cpu_dev, &pimpl->cpu_buft_list}; + } + const int layer_gpu = std::upper_bound(splits.begin(), splits.begin() + n_devices(), float(il - i_gpu_start)/act_gpu_layers) - splits.begin(); + auto * dev = devices.at(layer_gpu); + return {dev, &pimpl->gpu_buft_list.at(dev)}; + }; - vocab.n_vocab = n_vocab; - vocab.id_to_token.resize(n_vocab); + // assign the input layer + // there is very little benefit to offloading the input layer, so always keep it on the CPU + pimpl->dev_input = { cpu_dev, &pimpl->cpu_buft_list }; - for (uint32_t i = 0; i < n_vocab; i++) { - std::string word = gguf_get_arr_str(ctx, token_idx, i); - if (word.empty()) { - LLAMA_LOG_WARN("%s: empty token at index %u\n", __func__, i); - word = "[EMPTY_" + std::to_string(i) + "]"; - } + // assign the repeating layers to the devices according to the splits + pimpl->dev_layer.resize(n_layer); + for (int il = 0; il < n_layer; ++il) { + pimpl->dev_layer[il] = get_layer_buft_list(il); + } - vocab.token_to_id[word] = i; - vocab.max_token_len = std::max(vocab.max_token_len, (int) word.size()); - - auto & token_data = vocab.id_to_token[i]; - token_data.text = std::move(word); - token_data.score = scores ? scores[i] : 0.0f; - token_data.attr = LLAMA_TOKEN_ATTR_NORMAL; - - if (toktypes) { //TODO: remove, required until per token attributes are available from GGUF file - switch(toktypes[i]) { - case LLAMA_TOKEN_TYPE_UNKNOWN: token_data.attr = LLAMA_TOKEN_ATTR_UNKNOWN; break; - case LLAMA_TOKEN_TYPE_UNUSED: token_data.attr = LLAMA_TOKEN_ATTR_UNUSED; break; - case LLAMA_TOKEN_TYPE_NORMAL: token_data.attr = LLAMA_TOKEN_ATTR_NORMAL; break; - case LLAMA_TOKEN_TYPE_CONTROL: token_data.attr = LLAMA_TOKEN_ATTR_CONTROL; break; - case LLAMA_TOKEN_TYPE_USER_DEFINED: token_data.attr = LLAMA_TOKEN_ATTR_USER_DEFINED; break; - case LLAMA_TOKEN_TYPE_BYTE: token_data.attr = LLAMA_TOKEN_ATTR_BYTE; break; - case LLAMA_TOKEN_TYPE_UNDEFINED: token_data.attr = LLAMA_TOKEN_ATTR_UNDEFINED; break; - default: token_data.attr = LLAMA_TOKEN_ATTR_UNDEFINED; break; + // assign the output layer + pimpl->dev_output = get_layer_buft_list(n_layer); + + // one ggml context per buffer type + int max_n_tensors = ml.n_tensors; + max_n_tensors += 1; // duplicated output tensor + max_n_tensors += n_layer*2; // duplicated rope freq tensors + const size_t ctx_size = ggml_tensor_overhead()*max_n_tensors; + + std::map ctx_map; + auto ctx_for_buft = [&](ggml_backend_buffer_type_t buft) -> ggml_context * { + auto it = ctx_map.find(buft); + if (it == ctx_map.end()) { + ggml_init_params params = { + /*.mem_size =*/ ctx_size, + /*.mem_buffer =*/ NULL, + /*.no_alloc =*/ true, + }; + + ggml_context * ctx = ggml_init(params); + if (!ctx) { + throw std::runtime_error(format("failed to create ggml context")); } - } - } - GGML_ASSERT(vocab.id_to_token.size() == vocab.token_to_id.size()); - vocab.init_tokenizer(); + ctx_map[buft] = ctx; + pimpl->ctxs.emplace_back(ctx); - // determine the newline token: LLaMA "<0x0A>" == 10 == '\n', Falcon 193 == '\n' - if (vocab.type == LLAMA_VOCAB_TYPE_SPM) { - try { - vocab.linefeed_id = llama_byte_to_token_impl(vocab, '\n'); - } catch (const std::exception & e) { - LLAMA_LOG_WARN("%s: SPM vocabulary, but newline token not found: %s! Using special_pad_id instead.", __func__, e.what()); - vocab.linefeed_id = vocab.special_pad_id; + return ctx; } - } else if (vocab.type == LLAMA_VOCAB_TYPE_WPM) { - vocab.linefeed_id = vocab.special_pad_id; - } else if (vocab.type == LLAMA_VOCAB_TYPE_RWKV) { - const std::vector ids = llama_tokenize_internal(vocab, "\n", false); - GGML_ASSERT(!ids.empty() && "model vocab missing newline token"); - vocab.linefeed_id = ids[0]; - } else { - const std::vector ids = llama_tokenize_internal(vocab, "\xC4\x8A", false); // U+010A + return it->second; + }; - //GGML_ASSERT(!ids.empty() && "model vocab missing newline token"); - if (ids.empty()) { - LLAMA_LOG_WARN("%s: model vocab missing newline token, using special_pad_id instead\n", __func__); - vocab.linefeed_id = vocab.special_pad_id; - } else { - vocab.linefeed_id = ids[0]; - } - } + const auto TENSOR_DUPLICATED = llama_model_loader::TENSOR_DUPLICATED; + const auto TENSOR_NOT_REQUIRED = llama_model_loader::TENSOR_NOT_REQUIRED; - // special tokens + // create tensors for the weights { - const std::vector> special_token_types = { - { LLM_KV_TOKENIZER_BOS_ID, vocab.special_bos_id }, - { LLM_KV_TOKENIZER_EOS_ID, vocab.special_eos_id }, - { LLM_KV_TOKENIZER_EOT_ID, vocab.special_eot_id }, - { LLM_KV_TOKENIZER_EOM_ID, vocab.special_eom_id }, - { LLM_KV_TOKENIZER_UNK_ID, vocab.special_unk_id }, - { LLM_KV_TOKENIZER_SEP_ID, vocab.special_sep_id }, - { LLM_KV_TOKENIZER_PAD_ID, vocab.special_pad_id }, - { LLM_KV_TOKENIZER_CLS_ID, vocab.special_cls_id }, - { LLM_KV_TOKENIZER_MASK_ID, vocab.special_mask_id }, - { LLM_KV_TOKENIZER_FIM_PRE_ID, vocab.special_fim_pre_id }, - { LLM_KV_TOKENIZER_FIM_SUF_ID, vocab.special_fim_suf_id }, - { LLM_KV_TOKENIZER_FIM_MID_ID, vocab.special_fim_mid_id }, - { LLM_KV_TOKENIZER_FIM_PAD_ID, vocab.special_fim_pad_id }, - { LLM_KV_TOKENIZER_FIM_REP_ID, vocab.special_fim_rep_id }, - { LLM_KV_TOKENIZER_FIM_SEP_ID, vocab.special_fim_sep_id }, - - // deprecated - { LLM_KV_TOKENIZER_PREFIX_ID, vocab.special_fim_pre_id }, - { LLM_KV_TOKENIZER_SUFFIX_ID, vocab.special_fim_suf_id }, - { LLM_KV_TOKENIZER_MIDDLE_ID, vocab.special_fim_mid_id }, - }; + // note: cast to int64_t since we will use these for the tensor dimensions + const int64_t n_head = hparams.n_head(); + const int64_t n_head_kv = hparams.n_head_kv(); + const int64_t n_embd = hparams.n_embd; + const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(); + const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(); + const int64_t n_embd_head_k = hparams.n_embd_head_k; + const int64_t n_embd_head_v = hparams.n_embd_head_v; + const int64_t n_ff = hparams.n_ff(); + const int64_t n_embd_gqa = n_embd_v_gqa; + const int64_t n_vocab = vocab.n_tokens(); + const int64_t n_token_types = vocab.n_token_types(); + const int64_t n_rot = hparams.n_rot; + const int64_t n_expert = hparams.n_expert; + const int64_t n_expert_used = hparams.n_expert_used; + const int64_t n_ctx_train = hparams.n_ctx_train; + + if (n_expert > 0 && hparams.n_expert_used == 0) { + throw std::runtime_error("model has expert layers but no expert layers are used"); + } - for (const auto & it : special_token_types) { - const std::string & key = kv(std::get<0>(it)); - int32_t & id = std::get<1>(it); + int n_moved_tensors = 0; + ggml_tensor * first_moved_tensor = nullptr; + ggml_backend_buffer_type_t first_moved_from_buft = nullptr; + ggml_backend_buffer_type_t first_moved_to_buft = nullptr; - uint32_t new_id; - if (!ml.get_key(std::get<0>(it), new_id, false)) { - continue; - } - if (new_id >= vocab.id_to_token.size()) { - LLAMA_LOG_WARN("%s: bad special token: '%s' = %ud, using default id %d\n", - __func__, key.c_str(), new_id, id); - } else { - id = new_id; + auto create_tensor = [&](const LLM_TN_IMPL & tn, const std::initializer_list & ne, int flags) -> ggml_tensor * { + ggml_tensor * t_meta = ml.get_tensor_meta(tn.str().c_str()); + + if (!t_meta) { + if (flags & TENSOR_NOT_REQUIRED) { + return nullptr; + } + throw std::runtime_error(format("missing tensor '%s'", tn.str().c_str())); } - } - // Handle add_bos_token and add_eos_token - { - bool temp = true; + // some models use the token embedding tensor as the output, but since these are used in different layers and with different ops + // the tensor is duplicated + // to handle this, we check if the tensor is duplicated, and if so, we assume that it is being loaded as the output tensor + llm_tensor tn_tensor = tn.tensor; + if (tn.tensor == LLM_TENSOR_TOKEN_EMBD && flags & TENSOR_DUPLICATED) { + tn_tensor = LLM_TENSOR_OUTPUT; + } - if (ml.get_key(LLM_KV_TOKENIZER_ADD_BOS, temp, false)) { - vocab.tokenizer_add_bos = temp; + llm_tensor_info info; + try { + info = llm_tensor_info_for(tn_tensor); + } catch (const std::out_of_range & e) { + throw std::runtime_error(format("missing tensor info mapping for %s", tn.str().c_str())); } - if (ml.get_key(LLM_KV_TOKENIZER_ADD_EOS, temp, false)) { - vocab.tokenizer_add_eos = temp; + + // tensors with "bias" suffix are always used with GGML_OP_ADD + ggml_op op; + bool bias = tn.suffix != nullptr && strcmp(tn.suffix, "bias") == 0; + if (bias) { + op = GGML_OP_ADD; + } else { + op = info.op; } - } - // auto-detect special tokens by text - // TODO: convert scripts should provide these tokens through the KV metadata LLM_KV_TOKENIZER_... - // for now, we apply this workaround to find the tokens based on their text - - for (const auto & t : vocab.token_to_id) { - // find EOT token: "<|eot_id|>", "<|im_end|>", "", etc. - if (vocab.special_eot_id == LLAMA_TOKEN_NULL) { - if (false - || t.first == "<|eot_id|>" - || t.first == "<|im_end|>" - || t.first == "<|end|>" - || t.first == "" - || t.first == "<|endoftext|>" - || t.first == "" - || t.first == "<｜end▁of▁sentence｜>" // DeepSeek - ) { - vocab.special_eot_id = t.second; - if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) { - LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n", - __func__, t.second, t.first.c_str()); - vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL; - } + // sanity checks + if (info.layer == LLM_TENSOR_LAYER_INPUT || info.layer == LLM_TENSOR_LAYER_OUTPUT) { + if (tn.bid != -1) { + GGML_ABORT("input/output layer tensor %s used with a layer number", tn.str().c_str()); + } + } else { + if (tn.bid == -1) { + GGML_ABORT("repeating layer tensor %s used without a layer number", tn.str().c_str()); } } - // find EOM token: "<|eom_id|>" - if (vocab.special_eom_id == LLAMA_TOKEN_NULL) { - if (false - || t.first == "<|eom_id|>" - ) { - vocab.special_eom_id = t.second; - if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) { - LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n", - __func__, t.second, t.first.c_str()); - vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL; - } - } + // select the buffer type for this tensor + buft_list_t * buft_list; + switch (info.layer) { + case LLM_TENSOR_LAYER_INPUT: + buft_list = pimpl->dev_input.buft_list; + break; + case LLM_TENSOR_LAYER_OUTPUT: + buft_list = pimpl->dev_output.buft_list; + break; + case LLM_TENSOR_LAYER_REPEATING: + buft_list = pimpl->dev_layer.at(tn.bid).buft_list; + break; + default: + GGML_ABORT("invalid layer %d for tensor %s", info.layer, tn.str().c_str()); } - // find FIM_PRE token: "<|fim_prefix|>", "", "
", etc.
-            if (vocab.special_fim_pre_id == LLAMA_TOKEN_NULL) {
-                if (false
-                        || t.first == "<|fim_prefix|>"  // Qwen
-                        || t.first == ""
-                        || t.first == "<｜fim▁begin｜>" // DeepSeek
-                        || t.first == "
"
-                        ) {
-                    vocab.special_fim_pre_id = t.second;
-                    if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
-                        LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n",
-                                __func__, t.second, t.first.c_str());
-                        vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL;
-                    }
-                }
+            ggml_backend_buffer_type_t buft = select_weight_buft(hparams, t_meta, op, *buft_list);
+            if (!buft) {
+                throw std::runtime_error(format("failed to find a compatible buffer type for tensor %s", tn.str().c_str()));
             }
 
-            // find FIM_SUF token: "<|fim_suffix|>", "", "", etc.
-            if (vocab.special_fim_suf_id == LLAMA_TOKEN_NULL) {
-                if (false
-                        || t.first == "<|fim_suffix|>" // Qwen
-                        || t.first == ""
-                        || t.first == "<｜fim▁hole｜>" // DeepSeek
-                        || t.first == ""
-                        ) {
-                    vocab.special_fim_suf_id = t.second;
-                    if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
-                        LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n",
-                                __func__, t.second, t.first.c_str());
-                        vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL;
-                    }
-                }
+            // avoid using a host buffer when using mmap
+            auto * buft_dev = ggml_backend_buft_get_device(buft);
+            if (ml.use_mmap && buft_dev && buft == ggml_backend_dev_host_buffer_type(buft_dev)) {
+                auto * cpu_dev = ggml_backend_dev_by_type(GGML_BACKEND_DEVICE_TYPE_CPU);
+                buft = ggml_backend_dev_buffer_type(cpu_dev);
             }
 
-            // find FIM_MID token: "<|fim_middle|>", "", "", etc.
-            if (vocab.special_fim_mid_id == LLAMA_TOKEN_NULL) {
-                if (false
-                        || t.first == "<|fim_middle|>" // Qwen
-                        || t.first == ""
-                        || t.first == "<｜fim▁end｜>"  // DeepSeek
-                        || t.first == ""
-                        ) {
-                    vocab.special_fim_mid_id = t.second;
-                    if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
-                        LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n",
-                                __func__, t.second, t.first.c_str());
-                        vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL;
-                    }
+            if (buft != buft_list->front().second) {
+                n_moved_tensors++;
+                if (!first_moved_tensor) {
+                    first_moved_tensor = t_meta;
+                    first_moved_from_buft = buft_list->front().second;
+                    first_moved_to_buft   = buft;
                 }
             }
 
-            // find FIM_PAD token: "<|fim_pad|>", "", "", etc.
-            if (vocab.special_fim_pad_id == LLAMA_TOKEN_NULL) {
-                if (false
-                        || t.first == "<|fim_pad|>" // Qwen
-                        || t.first == ""
-                        || t.first == ""
-                        ) {
-                    vocab.special_fim_pad_id = t.second;
-                    if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
-                        LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n",
-                                __func__, t.second, t.first.c_str());
-                        vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL;
-                    }
+            ggml_context * ctx = ctx_for_buft(buft);
+
+            // if duplicated, check if the original tensor was allocated in the same buffer type context and avoid creating a new one
+            if (flags & TENSOR_DUPLICATED) {
+                ggml_tensor * t = ggml_get_tensor(ctx, tn.str().c_str());
+                if (t) {
+                    return t;
                 }
             }
+            return ml.create_tensor(ctx, tn, ne, flags);
+        };
 
-            // find FIM_REP token: "<|fim_repo|>", "", "", etc.
-            if (vocab.special_fim_rep_id == LLAMA_TOKEN_NULL) {
-                if (false
-                        || t.first == "<|fim_repo|>"  // Qwen
-                        || t.first == "<|repo_name|>"
-                        || t.first == ""
-                        || t.first == ""
-                        ) {
-                    vocab.special_fim_rep_id = t.second;
-                    if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
-                        LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n",
-                                __func__, t.second, t.first.c_str());
-                        vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL;
+        layers.resize(n_layer);
+
+        // TODO: move to a separate function
+        const auto tn = LLM_TN(arch);
+        switch (arch) {
+            case LLM_ARCH_LLAMA:
+            case LLM_ARCH_REFACT:
+            case LLM_ARCH_MINICPM:
+            case LLM_ARCH_GRANITE:
+            case LLM_ARCH_GRANITE_MOE:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
                     }
-                }
-            }
 
-            // find FIM_SEP token: "<|file_sep|>"
-            if (vocab.special_fim_sep_id == LLAMA_TOKEN_NULL) {
-                if (false
-                        || t.first == "<|file_sep|>" // Qwen
-                        ) {
-                    vocab.special_fim_sep_id = t.second;
-                    if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
-                        LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n",
-                                __func__, t.second, t.first.c_str());
-                        vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL;
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+                        // optional bias tensors
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_embd},     TENSOR_NOT_REQUIRED);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_embd_gqa}, TENSOR_NOT_REQUIRED);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_embd_gqa}, TENSOR_NOT_REQUIRED);
+                        layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd},     TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+                            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                        }
+                        else {
+                            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                        }
+
+                        if (n_expert == 0) {
+                            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+                            // optional MLP bias
+                            layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+                            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+                        } else {
+                            layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+                            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, TENSOR_NOT_REQUIRED);
+                            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+                            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+                        }
+                    }
+                } break;
+            case LLM_ARCH_DECI:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
                     }
-                }
-            }
-        }
 
-        // maintain a list of tokens that cause end-of-generation
-        // this is currently determined based on the token text, which is obviously not ideal
-        // ref: https://github.com/ggerganov/llama.cpp/issues/9606
-        vocab.special_eog_ids.clear();
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+                        const int64_t n_embd_k_gqa  = hparams.n_embd_k_gqa(i);
+                        const int64_t n_embd_v_gqa  = hparams.n_embd_v_gqa(i);
+                        const int64_t n_embd_gqa    = hparams.n_embd_v_gqa(i);
+                        const int64_t n_ff          = hparams.n_ff(i);
+                        const int64_t n_head        = hparams.n_head(i);
+                        const int64_t n_head_kv     = hparams.n_head_kv(i);
+
+                        if (n_head_kv == 0 && n_head > 0) {
+                            // linear attention for DeciLMCausalModel
+                            layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        }
+                        else if (n_head_kv > 0) {
+                            layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                            layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+                            layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                            layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                            layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+                        }
+
+                        // optional bias tensors
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_embd},     TENSOR_NOT_REQUIRED);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_embd_gqa}, TENSOR_NOT_REQUIRED);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_embd_gqa}, TENSOR_NOT_REQUIRED);
+                        layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd},     TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        if (hparams.rope_scaling_type_train == LLAMA_ROPE_SCALING_TYPE_LONGROPE) {
+                            layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                            layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                        }
+                        else {
+                            layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                        }
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+                        // optional MLP bias
+                        layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+                    }
+                } break;
+            case LLM_ARCH_MINICPM3:
+                {
+                    const int64_t n_embd_head_qk_rope = hparams.n_rot;
+                    const int64_t n_embd_head_qk_nope = hparams.n_embd_head_k - hparams.n_rot;
+
+                    const int64_t q_lora_rank  = hparams.n_lora_q;
+                    const int64_t kv_lora_rank = hparams.n_lora_kv;
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
 
-        if (vocab.special_fim_pad_id != LLAMA_TOKEN_NULL && vocab.special_eog_ids.count(vocab.special_fim_pad_id) == 0) {
-            vocab.special_eog_ids.insert(vocab.special_fim_pad_id);
-        }
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
 
-        if (vocab.special_fim_rep_id != LLAMA_TOKEN_NULL && vocab.special_eog_ids.count(vocab.special_fim_rep_id) == 0) {
-            vocab.special_eog_ids.insert(vocab.special_fim_rep_id);
-        }
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_q_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_A_NORM, "weight", i), {q_lora_rank}, 0);
 
-        if (vocab.special_fim_sep_id != LLAMA_TOKEN_NULL && vocab.special_eog_ids.count(vocab.special_fim_sep_id) == 0) {
-            vocab.special_eog_ids.insert(vocab.special_fim_sep_id);
-        }
+                        layer.attn_kv_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_NORM, "weight", i), {kv_lora_rank}, 0);
 
-        for (const auto & t : vocab.token_to_id) {
-            if (false
-                    || t.first == "<|eot_id|>"
-                    || t.first == "<|im_end|>"
-                    || t.first == "<|end|>"
-                    || t.first == ""
-                    || t.first == "<|endoftext|>"
-                    || t.first == "<|eom_id|>"
-                    || t.first == ""
-               ) {
-                vocab.special_eog_ids.insert(t.second);
-                if ((vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
-                    LLAMA_LOG_WARN("%s: control-looking token: %6d '%s' was not control-type; this is probably a bug in the model. its type will be overridden\n",
-                            __func__, t.second, t.first.c_str());
-                    vocab.id_to_token[t.second].attr = LLAMA_TOKEN_ATTR_CONTROL;
-                }
-            } else {
-                // token is control, but not marked as EOG -> print a debug log
-                if (vocab.id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL && vocab.special_eog_ids.count(t.second) == 0) {
-                    LLAMA_LOG_DEBUG("%s: control token: %6d '%s' is not marked as EOG\n",
-                            __func__, t.second, t.first.c_str());
-                }
-            }
-        }
+                        layer.wq_a = create_tensor(tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}, 0);
+                        layer.wq_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, n_head * n_embd_head_k}, 0);
 
-        // sanity checks
-        if (vocab.special_eos_id != LLAMA_TOKEN_NULL && vocab.special_eog_ids.count(vocab.special_eos_id) == 0) {
-            vocab.special_eog_ids.insert(vocab.special_eos_id);
-            LLAMA_LOG_WARN("%s: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect\n", __func__);
-        }
+                        layer.wkv_a_mqa = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + (n_embd_head_qk_rope)}, 0);
+                        layer.wkv_b     = create_tensor(tn(LLM_TENSOR_ATTN_KV_B,     "weight", i), {kv_lora_rank, n_head * (n_embd_head_qk_nope + n_embd_head_v)}, 0);
+                        layer.wo        = create_tensor(tn(LLM_TENSOR_ATTN_OUT,      "weight", i), {              n_head * (                      n_embd_head_v), n_embd}, 0);
 
-        if (vocab.special_eot_id != LLAMA_TOKEN_NULL && vocab.special_eog_ids.count(vocab.special_eot_id) == 0) {
-            vocab.special_eog_ids.insert(vocab.special_eot_id);
-            LLAMA_LOG_WARN("%s: special_eot_id is not in special_eog_ids - the tokenizer config may be incorrect\n", __func__);
-        }
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
 
-        if (vocab.special_eom_id != LLAMA_TOKEN_NULL && vocab.special_eog_ids.count(vocab.special_eom_id) == 0) {
-            vocab.special_eog_ids.insert(vocab.special_eom_id);
-            LLAMA_LOG_WARN("%s: special_eom_id is not in special_eog_ids - the tokenizer config may be incorrect\n", __func__);
-        }
-    }
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
 
-    // build special tokens cache
-    {
-        for (llama_vocab::id id = 0; id < (llama_vocab::id)n_vocab; ++id) {
-            if (vocab.id_to_token[id].attr & (LLAMA_TOKEN_ATTR_CONTROL | LLAMA_TOKEN_ATTR_USER_DEFINED | LLAMA_TOKEN_ATTR_UNKNOWN)) {
-                vocab.cache_special_tokens.push_back(id);
-            }
-        }
+                        layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), { n_embd_head_qk_rope/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                        layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), { n_embd_head_qk_rope/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                    }
+                } break;
+            case LLM_ARCH_GROK:
+                {
+                    if (n_expert == 0) {
+                        throw std::runtime_error("Grok model cannot have zero experts");
+                    }
 
-        std::sort(vocab.cache_special_tokens.begin(), vocab.cache_special_tokens.end(),
-            [&] (const llama_vocab::id a, const llama_vocab::id b) {
-                return vocab.id_to_token[a].text.size() > vocab.id_to_token[b].text.size();
-            }
-        );
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
 
-        LLAMA_LOG_INFO("%s: special tokens cache size = %u\n", __func__, (uint32_t)vocab.cache_special_tokens.size());
-    }
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
 
-    // build token to piece cache
-    {
-        size_t size_cache = 0;
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
 
-        std::vector cache_token_to_piece(n_vocab);
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
 
-        for (uint32_t id = 0; id < n_vocab; ++id) {
-            cache_token_to_piece[id] = llama_token_to_piece(&model, id, true);
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
 
-            size_cache += cache_token_to_piece[id].size();
-        }
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
 
-        std::swap(vocab.cache_token_to_piece, cache_token_to_piece);
+                        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
 
-        LLAMA_LOG_INFO("%s: token to piece cache size = %.4f MB\n", __func__, size_cache / 1024.0 / 1024.0);
-    }
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
 
-    // Handle per token attributes
-    //NOTE: Each model customizes per token attributes.
-    //NOTE: Per token attributes are missing from the GGUF file.
-    //TODO: Extract attributes from GGUF file.
-    {
-        auto _contains_any = [] (const std::string &str, const std::vector &substrs) -> bool {
-            for (auto substr : substrs) {
-                if (str.find(substr) < std::string::npos) {
-                    return true;
-                }
-            }
-            return false;
-        };
+                        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+                        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff, n_expert}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+                        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
 
-        auto _set_tokenid_attr = [&] (const llama_vocab::id id, llama_token_attr attr, bool value) {
-            uint32_t current = vocab.id_to_token.at(id).attr;
-            current = value ? (current | attr) : (current & ~attr);
-            vocab.id_to_token[id].attr = (llama_token_attr) current;
-        };
+                        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+                    }
+                } break;
+            case LLM_ARCH_DBRX:
+                {
+                    if (n_expert == 0) {
+                        throw std::runtime_error("DBRX model cannot have zero experts");
+                    }
 
-        auto _set_token_attr = [&] (const std::string & token, llama_token_attr attr, bool value) {
-            _set_tokenid_attr(vocab.token_to_id.at(token), attr, value);
-        };
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
 
-        std::string model_name;
-        std::string tokenizer_pre;
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
 
-        ml.get_key(LLM_KV_GENERAL_NAME, model_name, false);
-        ml.get_key(LLM_KV_TOKENIZER_PRE, tokenizer_pre, false);
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
 
-        // model name to lowercase
-        std::transform(model_name.begin(), model_name.end(), model_name.begin(),
-            [] (const std::string::value_type x) {
-                return std::tolower(x);
-            }
-        );
-
-        // set attributes by model/tokenizer name
-        if (_contains_any(tokenizer_pre, {"jina-v2-de", "jina-v2-es", "jina-v2-code"})) {
-            _set_token_attr("", LLAMA_TOKEN_ATTR_LSTRIP, true);
-        } else if (_contains_any(model_name, {"phi-3", "phi3"})) {
-            for (auto id : vocab.cache_special_tokens) {
-                _set_tokenid_attr(id, LLAMA_TOKEN_ATTR_RSTRIP, true);
-            }
-            for (auto token : {""}) {
-                _set_token_attr(token, LLAMA_TOKEN_ATTR_RSTRIP, true);
-            }
-            for (auto token : {"", "", "<|endoftext|>"}) {
-                _set_token_attr(token, LLAMA_TOKEN_ATTR_RSTRIP, false);
-            }
-        }
-    }
-}
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
 
-void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
-    const auto & hparams = model.hparams;
-    const auto & vocab   = model.vocab;
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
 
-    const char * rope_scaling_type = LLAMA_ROPE_SCALING_TYPES.at(hparams.rope_scaling_type_train);
+                        layer.attn_out_norm = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
 
-    auto print_f = [](const std::function & f, uint32_t n) {
-        bool is_var = false;
+                        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {n_embd, n_expert}, 0);
+                        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff,   n_expert}, 0);
+                        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff,   n_embd, n_expert}, 0);
+                        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff,   n_expert}, 0);
+                    }
+                } break;
+            case LLM_ARCH_BAICHUAN:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+                    {
+                        output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                        output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+                    }
 
-        std::vector v;
-        for (uint32_t i = 0; i < n; ++i) {
-            v.push_back(f(i));
-            if (v[i] != v[0]) {
-                is_var = true;
-            }
-        }
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
 
-        std::stringstream ss;
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
 
-        if (is_var) {
-            ss << "[";
-            for (uint32_t i = 0; i < n; ++i) {
-                ss << v[i];
-                if (i < n - 1) {
-                    ss << ", ";
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_FALCON:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    {
+                        output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                        output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+
+                        output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                        if (!output) {
+                            output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); // needs to be on GPU
+                        }
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.attn_norm_2   = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.attn_norm_2_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_STARCODER:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+                    pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,   "weight"), {n_embd, n_ctx_train}, 0);
+
+                    // output
+                    {
+                        output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                        output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                        output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                        if (!output) {
+                            // needs to be on GPU
+                            output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                        }
+
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa}, 0);
+
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i),   {n_embd, n_ff}, 0);
+                        layer.ffn_up_b = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i),     {n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_BERT:
+            case LLM_ARCH_NOMIC_BERT:
+                {
+                    tok_embd     = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, 0);
+                    type_embd    = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, 0);
+
+                    if (arch == LLM_ARCH_BERT) {
+                        pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,    "weight"), {n_embd, n_ctx_train}, 0);
+
+                        cls   = create_tensor(tn(LLM_TENSOR_CLS, "weight"), {n_embd, n_embd}, TENSOR_NOT_REQUIRED);
+                        cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"),   {n_embd},         TENSOR_NOT_REQUIRED);
+
+                        cls_out   = create_tensor(tn(LLM_TENSOR_CLS_OUT, "weight"), {n_embd, 1}, TENSOR_NOT_REQUIRED);
+                        cls_out_b = create_tensor(tn(LLM_TENSOR_CLS_OUT, "bias"),   {1},         TENSOR_NOT_REQUIRED);
+                    }
+
+                    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd}, 0);
+                    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"),   {n_embd}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        if (arch == LLM_ARCH_BERT) {
+                            layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                            layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i),   {n_embd}, 0);
+
+                            layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                            layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i),   {n_embd_gqa}, 0);
+
+                            layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                            layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i),   {n_embd_gqa}, 0);
+                        } else {
+                            layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        }
+
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT,      "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_out_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,        "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN,      "weight", i), {n_ff, n_embd}, 0);
+
+                        if (arch == LLM_ARCH_BERT) {
+                            layer.bo         = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0);
+                            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, 0);
+                            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, 0);
+                        } else {
+                            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+                        }
+
+                        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+                        layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias", i),   {n_embd}, 0);
+                    }
+                } break;
+            case LLM_ARCH_JINA_BERT_V2:
+                {
+                    tok_embd  = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, 0); // word_embeddings
+                    type_embd = create_tensor(tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_token_types}, 0); // token_type_embeddings
+
+                    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd}, 0); // LayerNorm
+                    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"),   {n_embd}, 0); //LayerNorm bias
+
+                    cls   = create_tensor(tn(LLM_TENSOR_CLS, "weight"), {n_embd, 1}, TENSOR_NOT_REQUIRED);
+                    cls_b = create_tensor(tn(LLM_TENSOR_CLS, "bias"),   {1},         TENSOR_NOT_REQUIRED);
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i]; // JinaBertLayer
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "bias", i),   {n_embd}, 0);
+
+                        layer.attn_q_norm   = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.attn_q_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K, "bias",   i), {n_embd_gqa}, 0);
+
+                        layer.attn_k_norm   = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.attn_k_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V, "bias",   i), {n_embd_gqa}, 0);
+
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0); //output_dens
+                        layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias",   i), {n_embd}, 0); //output_dens
+
+                        layer.attn_out_norm   = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}, 0); //output_norm
+                        layer.attn_out_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_OUT_NORM, "bias",   i), {n_embd}, 0);
+
+                        layer.attn_norm_2   = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.attn_norm_2_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias",   i), {n_embd}, 0);
+
+                        layer.layer_out_norm   = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}, 0);
+                        layer.layer_out_norm_b = create_tensor(tn(LLM_TENSOR_LAYER_OUT_NORM, "bias",   i), {n_embd}, 0);
+                    }
+                } break;
+            case LLM_ARCH_BLOOM:
+                {
+                    tok_embd   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,      "weight"), {n_embd, n_vocab}, 0);
+                    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd}, 0);
+                    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"),   {n_embd}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias",   i), {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias",   i), {n_embd + 2*n_embd_gqa}, 0);
+
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias",   i), {n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias",   i), {n_embd}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias",   i), {n_embd}, 0);
+
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias",   i), {n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_MPT:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+                    pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,   "weight"), {n_embd, n_ctx_train}, TENSOR_NOT_REQUIRED);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                    if (!output) {
+                        output    = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); // needs to be on GPU
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa}, TENSOR_NOT_REQUIRED);
+
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, TENSOR_NOT_REQUIRED);
+
+                        layer.attn_q_norm   = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.attn_q_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.attn_k_norm   = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.attn_k_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        // AWQ ScaleActivation layer
+                        layer.ffn_act = create_tensor(tn(LLM_TENSOR_FFN_ACT, "scales", i), {n_ff}, TENSOR_NOT_REQUIRED);
+                    }
+                } break;
+            case LLM_ARCH_STABLELM:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm =   create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        // optional bias tensors, present in Stable LM 2 1.6B
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_embd},     TENSOR_NOT_REQUIRED);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_embd_gqa}, TENSOR_NOT_REQUIRED);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_embd_gqa}, TENSOR_NOT_REQUIRED);
+
+                        // optional q and k layernorms, present in StableLM 2 12B
+                        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k, n_head},    TENSOR_NOT_REQUIRED);
+                        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k, n_head_kv}, TENSOR_NOT_REQUIRED);
+
+                        // optional FFN norm, not present in StableLM 2 12B which uses parallel residual
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_QWEN:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd*3}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd*3}, 0);
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff/2}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff/2, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff/2}, 0);
+                    }
+                } break;
+            case LLM_ARCH_QWEN2:
+            case LLM_ARCH_QWEN2VL:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        // optional bias tensors
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_embd}, 0);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_embd_gqa}, 0);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_embd_gqa}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_QWEN2MOE:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        // optional bias tensors
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_embd}, 0);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_embd_gqa}, 0);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_embd_gqa}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+                        if (n_expert == 0) {
+                            throw std::runtime_error("n_expert must be > 0 for QWEN2MOE");
+                        }
+                        if (n_expert_used == 0) {
+                            throw std::runtime_error("n_expert_used must be > 0 for QWEN2MOE");
+                        }
+
+                        // MoE branch
+                        const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+
+                        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+                        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+                        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+                        // Shared expert branch
+                        const int64_t n_ff_shexp = hparams.n_ff_shexp ? hparams.n_ff_shexp : n_ff;
+
+                        layer.ffn_gate_inp_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP_SHEXP, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {    n_embd, n_ff_shexp}, 0);
+                        layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp,     n_embd}, 0);
+                        layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {    n_embd, n_ff_shexp}, 0);
+                    }
+                } break;
+            case LLM_ARCH_PHI2:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+                    output_b      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "bias"),   {n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, TENSOR_NOT_REQUIRED);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa}, TENSOR_NOT_REQUIRED);
+
+                        if (layer.wqkv == nullptr) {
+                            layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}, 0);
+                            layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "bias", i),   {n_embd}, 0);
+
+                            layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}, 0);
+                            layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K, "bias", i),   {n_embd_gqa}, 0);
+
+                            layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}, 0);
+                            layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V, "bias", i),   {n_embd_gqa}, 0);
+                        }
+
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_PHI3:
+                {
+                    const int64_t n_embd_head = n_embd / n_head;
+
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), { n_embd, n_vocab }, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), { n_embd, n_embd + 2 * n_embd_gqa }, TENSOR_NOT_REQUIRED);
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd, n_embd }, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), { n_embd }, 0);
+
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, 0);
+                        layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), { n_embd, 2 * n_ff }, 0);
+
+                        layer.rope_long  = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_LONG,  "weight", i), { n_embd_head/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                        layer.rope_short = create_tensor(tn(LLM_TENSOR_ROPE_FACTORS_SHORT, "weight", i), { n_embd_head/2 }, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                    }
+                } break;
+            case LLM_ARCH_PLAMO:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_GPT2:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+                    pos_embd = create_tensor(tn(LLM_TENSOR_POS_EMBD,   "weight"), {n_embd, n_ctx_train}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "bias", i),   {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa}, 0);
+
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_CODESHELL:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa}, 0);
+
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i),   {n_embd, n_ff}, 0);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP, "bias", i),     {n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_ORION:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_INTERNLM2:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        // layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_GEMMA:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); // same as tok_embd, duplicated to allow offloading
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                    }
+                } break;
+            case LLM_ARCH_GEMMA2:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,  "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); // same as tok_embd, duplicated to allow offloading
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+                        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+                    }
+                } break;
+            case LLM_ARCH_STARCODER2:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        // optional bias tensors
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_embd}, 0);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_embd_gqa}, 0);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_embd_gqa}, 0);
+                        layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+                        // optional bias tensors
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, 0);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP ,  "bias", i), {  n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_MAMBA:
+                {
+                    const int64_t d_conv  = hparams.ssm_d_conv;
+                    const int64_t d_inner = hparams.ssm_d_inner;
+                    const int64_t d_state = hparams.ssm_d_state;
+                    const int64_t dt_rank = hparams.ssm_dt_rank;
+
+                    // only an expansion factor of 2 is supported for now
+                    if (2 * n_embd != d_inner) {
+                        throw std::runtime_error("only an expansion factor of 2 is supported for now");
+                    }
+
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                    // if output is NULL, init from the input tok embed, duplicated to allow offloading
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        // norm
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ssm_in = create_tensor(tn(LLM_TENSOR_SSM_IN, "weight", i), {n_embd, 2*d_inner}, 0);
+
+                        layer.ssm_conv1d = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "weight", i), {d_conv, d_inner}, 0);
+                        layer.ssm_conv1d_b = create_tensor(tn(LLM_TENSOR_SSM_CONV1D, "bias", i), {d_inner}, 0);
+
+                        layer.ssm_x = create_tensor(tn(LLM_TENSOR_SSM_X, "weight", i), {d_inner, dt_rank + 2*d_state}, 0);
+
+                        layer.ssm_dt = create_tensor(tn(LLM_TENSOR_SSM_DT, "weight", i), {dt_rank, d_inner}, 0);
+                        layer.ssm_dt_b = create_tensor(tn(LLM_TENSOR_SSM_DT, "bias", i), {d_inner}, 0);
+
+                        // no "weight" suffix for these
+                        layer.ssm_a = create_tensor(tn(LLM_TENSOR_SSM_A, i), {d_state, d_inner}, 0);
+                        layer.ssm_d = create_tensor(tn(LLM_TENSOR_SSM_D, i), {d_inner}, 0);
+
+                        // out_proj
+                        layer.ssm_out = create_tensor(tn(LLM_TENSOR_SSM_OUT, "weight", i), {d_inner, n_embd}, 0);
+                    }
+                } break;
+            case LLM_ARCH_XVERSE:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_COMMAND_R:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    // init output from the input tok embed
+                    output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        if (n_layer >= 64){
+                            layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k, n_head}, 0);
+                            layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k, n_head_kv}, 0);
+                        }
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_COHERE2:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+                    // init output from the input tok embed
+                    output      = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab },
+                                                      TENSOR_DUPLICATED);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), { n_embd, n_embd }, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), { n_embd, n_embd_gqa }, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), { n_embd, n_embd_gqa }, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd, n_embd }, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), { n_embd, n_ff }, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), { n_embd, n_ff }, 0);
+                    }
+                }
+                break;
+            case LLM_ARCH_OLMO:  // adapted from LLM_ARCH_LLAMA with norm params removed
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_OLMO2:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+                    }
+                } break;
+            case LLM_ARCH_OLMOE:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+                        if (n_expert == 0) {
+                            throw std::runtime_error("n_expert must be > 0");
+                        }
+                        if (n_expert_used == 0) {
+                            throw std::runtime_error("n_expert_used must be > 0");
+                        }
+
+                        // MoE branch
+                        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff,   n_expert}, 0);
+                        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff,   n_embd, n_expert}, 0);
+                        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff,   n_expert}, 0);
+                    }
+                } break;
+            case LLM_ARCH_OPENELM:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    // init output from the input tok embed
+                    output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        const int64_t n_head      =   hparams.n_head(i);
+                        const int64_t n_head_qkv  = 2*hparams.n_head_kv(i) + n_head;
+                        const int64_t n_ff        =   hparams.n_ff(i);
+
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_head_qkv*n_embd_head_k}, 0);
+                        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+                        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head*n_embd_head_k, n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_GPTNEOX:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa}, 0);
+
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_ARCTIC:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_embd}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_embd, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+                        layer.ffn_norm_exps = create_tensor(tn(LLM_TENSOR_FFN_NORM_EXPS, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd,   n_ff, n_expert}, false);
+                        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {  n_ff, n_embd, n_expert}, 0);
+                        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd,   n_ff, n_expert}, 0);
+                    }
+                } break;
+            case LLM_ARCH_DEEPSEEK:
+                {
+
+                    const int64_t n_ff_exp        = hparams.n_ff_exp;
+                    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        if (i < (int) hparams.n_layer_dense_lead) {
+                            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                        } else {
+                            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+
+                            if (n_expert == 0) {
+                                throw std::runtime_error("n_expert must be > 0");
+                            }
+                            if (n_expert_used == 0) {
+                                throw std::runtime_error("n_expert_used must be > 0");
+                            }
+
+                            // MoE branch
+                            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+                            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+                            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+                            // Shared expert branch
+                            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+                            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, 0);
+                            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+                        }
+                    }
+                } break;
+            case LLM_ARCH_DEEPSEEK2:
+                {
+                    const bool is_lite = (hparams.n_layer == 27);
+
+                    const int64_t n_embd_head_qk_rope = hparams.n_rot;
+                    const int64_t n_embd_head_qk_nope = hparams.n_embd_head_k - hparams.n_rot;
+
+                    const int64_t q_lora_rank  = hparams.n_lora_q;
+                    const int64_t kv_lora_rank = hparams.n_lora_kv;
+
+                    const int64_t n_ff_exp        = hparams.n_ff_exp;
+                    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        if (!is_lite) {
+                            layer.attn_q_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_A_NORM, "weight", i), {q_lora_rank}, 0);
+                        }
+
+                        layer.attn_kv_a_norm = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_NORM, "weight", i), {kv_lora_rank}, 0);
+
+                        if (!is_lite) {
+                            layer.wq_a = create_tensor(tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}, 0);
+                            layer.wq_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, n_head * n_embd_head_k}, 0);
+                        } else {
+                            layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        }
+
+                        layer.wkv_a_mqa = create_tensor(tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + (n_embd_head_qk_rope)}, 0);
+                        layer.wkv_b     = create_tensor(tn(LLM_TENSOR_ATTN_KV_B,     "weight", i), {kv_lora_rank, n_head * (n_embd_head_qk_nope + n_embd_head_v)}, 0);
+                        layer.wo        = create_tensor(tn(LLM_TENSOR_ATTN_OUT,      "weight", i), {              n_head * (                      n_embd_head_v), n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        if (i < (int) hparams.n_layer_dense_lead) {
+                            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                        } else {
+                            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+                            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, TENSOR_NOT_REQUIRED);
+
+                            if (n_expert == 0) {
+                                throw std::runtime_error("n_expert must be > 0");
+                            }
+                            if (n_expert_used == 0) {
+                                throw std::runtime_error("n_expert_used must be > 0");
+                            }
+
+                            // MoE branch
+                            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+                            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+                            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+                            // Shared expert branch
+                            layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+                            layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {        n_ff_exp * n_expert_shared, n_embd}, 0);
+                            layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_exp * n_expert_shared}, 0);
+                        }
+                    }
+                } break;
+            case LLM_ARCH_BITNET:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm     = create_tensor(tn(LLM_TENSOR_ATTN_NORM,     "weight", i), {n_embd}, 0);
+                        layer.attn_sub_norm = create_tensor(tn(LLM_TENSOR_ATTN_SUB_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq       = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wq_scale = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+                        layer.wk       = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wk_scale = create_tensor(tn(LLM_TENSOR_ATTN_K,   "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+                        layer.wv       = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv_scale = create_tensor(tn(LLM_TENSOR_ATTN_V,   "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+                        layer.wo       = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.wo_scale = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_norm     = create_tensor(tn(LLM_TENSOR_FFN_NORM,     "weight", i), {n_embd}, 0);
+                        layer.ffn_sub_norm = create_tensor(tn(LLM_TENSOR_FFN_SUB_NORM, "weight", i), {n_ff}, 0);
+
+                        layer.ffn_gate       = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_gate_scale = create_tensor(tn(LLM_TENSOR_FFN_GATE, "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_down       = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_scale = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_up         = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_up_scale   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "scale",  i), {1}, TENSOR_NOT_REQUIRED);
+                    }
+                } break;
+            case LLM_ARCH_T5:
+                {
+                    const auto n_rel_attn_bkts = hparams.n_rel_attn_bkts;
+
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm_enc = create_tensor(tn(LLM_TENSOR_ENC_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm     = create_tensor(tn(LLM_TENSOR_DEC_OUTPUT_NORM, "weight"), {n_embd}, 0);
+
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm_enc  = create_tensor(tn(LLM_TENSOR_ENC_ATTN_NORM,  "weight", i), {n_embd}, 0);
+                        layer.attn_rel_b_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_REL_B, "weight", i), {n_head, n_rel_attn_bkts}, TENSOR_NOT_REQUIRED);
+
+                        layer.wq_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_Q,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wk_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_OUT, "weight", i), {n_embd_v_gqa, n_embd}, 0);
+
+                        layer.ffn_norm_enc = create_tensor(tn(LLM_TENSOR_ENC_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate_enc = create_tensor(tn(LLM_TENSOR_ENC_FFN_GATE, "weight", i), {n_embd,   n_ff}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_down_enc = create_tensor(tn(LLM_TENSOR_ENC_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up_enc   = create_tensor(tn(LLM_TENSOR_ENC_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+                        layer.attn_norm  = create_tensor(tn(LLM_TENSOR_DEC_ATTN_NORM,  "weight", i), {n_embd}, 0);
+                        layer.attn_rel_b = create_tensor(tn(LLM_TENSOR_DEC_ATTN_REL_B, "weight", i), {n_head, n_rel_attn_bkts}, TENSOR_NOT_REQUIRED);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_DEC_ATTN_Q,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_DEC_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_DEC_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_DEC_ATTN_OUT, "weight", i), {n_embd_v_gqa, n_embd}, 0);
+
+                        layer.attn_norm_cross  = create_tensor(tn(LLM_TENSOR_DEC_CROSS_ATTN_NORM,  "weight", i), {n_embd}, 0);
+                        // this tensor seems to be unused in HF transformers implementation
+                        layer.attn_rel_b_cross = create_tensor(tn(LLM_TENSOR_DEC_CROSS_ATTN_REL_B, "weight", i), {n_head, n_rel_attn_bkts}, TENSOR_NOT_REQUIRED);
+
+                        layer.wq_cross = create_tensor(tn(LLM_TENSOR_DEC_CROSS_ATTN_Q,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wk_cross = create_tensor(tn(LLM_TENSOR_DEC_CROSS_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv_cross = create_tensor(tn(LLM_TENSOR_DEC_CROSS_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo_cross = create_tensor(tn(LLM_TENSOR_DEC_CROSS_ATTN_OUT, "weight", i), {n_embd_v_gqa, n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_DEC_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_DEC_FFN_GATE, "weight", i), {n_embd,   n_ff}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_DEC_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_DEC_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_T5ENCODER:
+                {
+                    const auto n_rel_attn_bkts = hparams.n_rel_attn_bkts;
+
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm_enc = create_tensor(tn(LLM_TENSOR_ENC_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm_enc  = create_tensor(tn(LLM_TENSOR_ENC_ATTN_NORM,  "weight", i), {n_embd}, 0);
+                        layer.attn_rel_b_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_REL_B, "weight", i), {n_head, n_rel_attn_bkts}, TENSOR_NOT_REQUIRED);
+
+                        layer.wq_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_Q,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wk_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo_enc = create_tensor(tn(LLM_TENSOR_ENC_ATTN_OUT, "weight", i), {n_embd_v_gqa, n_embd}, 0);
+
+                        layer.ffn_norm_enc = create_tensor(tn(LLM_TENSOR_ENC_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_gate_enc = create_tensor(tn(LLM_TENSOR_ENC_FFN_GATE, "weight", i), {n_embd,   n_ff}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_down_enc = create_tensor(tn(LLM_TENSOR_ENC_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up_enc   = create_tensor(tn(LLM_TENSOR_ENC_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_JAIS:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM,   "bias", i),   {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa}, 0);
+
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+                        layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i),   {n_embd}, 0);
+
+                        layer.ffn_gate   = create_tensor(tn(LLM_TENSOR_FFN_GATE,   "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_gate_b = create_tensor(tn(LLM_TENSOR_FFN_GATE,   "bias", i),   {n_ff}, 0);
+
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i),   {n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_CHATGLM:
+                {
+                    tok_embd   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD,      "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, 0);
+                        layer.bqkv = create_tensor(tn(LLM_TENSOR_ATTN_QKV, "bias", i),   {n_embd + 2*n_embd_gqa}, 0);
+
+                        layer.wo   = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff * 2}, 0);
+
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                    }
+                } break;
+            case LLM_ARCH_NEMOTRON:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}, 0);
+                    output        = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        // optional bias tensors
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_embd},     TENSOR_NOT_REQUIRED);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_embd_gqa}, TENSOR_NOT_REQUIRED);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_embd_gqa}, TENSOR_NOT_REQUIRED);
+                        layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd},     TENSOR_NOT_REQUIRED);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+                        layer.ffn_norm_b = create_tensor(tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}, 0);
+
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+
+                        // optional MLP bias
+                        layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, TENSOR_NOT_REQUIRED);
+                        layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias", i), {n_ff}, TENSOR_NOT_REQUIRED);
+                    }
+                } break;
+            case LLM_ARCH_EXAONE:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+                        layer.ffn_norm   = create_tensor(tn(LLM_TENSOR_FFN_NORM,   "weight", i), {n_embd}, 0);
+                        layer.rope_freqs = create_tensor(tn(LLM_TENSOR_ROPE_FREQS, "weight", i), {n_rot/2}, TENSOR_NOT_REQUIRED | (i != 0 ? TENSOR_DUPLICATED : 0));
+                        layer.ffn_gate   = create_tensor(tn(LLM_TENSOR_FFN_GATE,   "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN,   "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,     "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_RWKV6:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // Block 0, LN0
+                    tok_norm = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd}, 0);
+                    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"), {n_embd}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}, 0);
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, 0);
+
+                    const int time_mix_extra_dim = hparams.time_mix_extra_dim;
+                    const int time_decay_extra_dim = hparams.time_decay_extra_dim;
+                    const int head_size = hparams.wkv_head_size;
+                    const int attn_hidden_size = n_embd;
+                    const int ffn_size = hparams.n_ff_arr[0];
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "bias", i),   {n_embd}, 0);
+
+                        layer.attn_norm_2   = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "weight", i), {n_embd}, 0);
+                        layer.attn_norm_2_b = create_tensor(tn(LLM_TENSOR_ATTN_NORM_2, "bias", i),   {n_embd}, 0);
+
+                        layer.time_mix_w1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_W1, "weight", i), {n_embd, time_mix_extra_dim * 5}, 0);
+                        layer.time_mix_w2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_W2, "weight", i), {time_mix_extra_dim, n_embd, 5}, 0);
+
+                        layer.time_mix_lerp_x = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_X, "weight", i), {n_embd, 1, 1}, 0);
+                        layer.time_mix_lerp_w = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_W, "weight", i), {n_embd, 1, 1}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.time_mix_lerp_k = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_K, "weight", i), {n_embd, 1, 1}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.time_mix_lerp_v = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_V, "weight", i), {n_embd, 1, 1}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.time_mix_lerp_r = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_R, "weight", i), {n_embd, 1, 1}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.time_mix_lerp_g = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_G, "weight", i), {n_embd, 1, 1}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.time_mix_lerp_fused = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_FUSED, "weight", i), {n_embd, 1, 1, 5}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        GGML_ASSERT(!(layer.time_mix_lerp_fused == NULL && layer.time_mix_lerp_w == NULL));
+
+                        layer.time_mix_first = create_tensor(tn(LLM_TENSOR_TIME_MIX_FIRST, "weight", i), {head_size, n_embd / head_size}, 0);
+                        layer.time_mix_decay = create_tensor(tn(LLM_TENSOR_TIME_MIX_DECAY, "weight", i), {n_embd}, 0);
+                        layer.time_mix_decay_w1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_DECAY_W1, "weight", i), {n_embd, time_decay_extra_dim}, 0);
+                        layer.time_mix_decay_w2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_DECAY_W2, "weight", i), {time_decay_extra_dim, attn_hidden_size}, 0);
+                        layer.time_mix_key = create_tensor(tn(LLM_TENSOR_TIME_MIX_KEY, "weight", i), {attn_hidden_size, n_embd}, 0);
+                        layer.time_mix_value = create_tensor(tn(LLM_TENSOR_TIME_MIX_VALUE, "weight", i), {attn_hidden_size, n_embd}, 0);
+                        layer.time_mix_receptance = create_tensor(tn(LLM_TENSOR_TIME_MIX_RECEPTANCE, "weight", i), {attn_hidden_size, n_embd}, 0);
+                        layer.time_mix_gate = create_tensor(tn(LLM_TENSOR_TIME_MIX_GATE, "weight", i), {attn_hidden_size, n_embd}, 0);
+
+                        layer.time_mix_ln = create_tensor(tn(LLM_TENSOR_TIME_MIX_LN, "weight", i), {n_embd}, 0);
+                        layer.time_mix_ln_b = create_tensor(tn(LLM_TENSOR_TIME_MIX_LN, "bias", i), {n_embd}, 0);
+                        layer.time_mix_output = create_tensor(tn(LLM_TENSOR_TIME_MIX_OUTPUT, "weight", i), {n_embd, attn_hidden_size}, 0);
+
+                        layer.channel_mix_lerp_k = create_tensor(tn(LLM_TENSOR_CHANNEL_MIX_LERP_K, "weight", i), {n_embd, 1, 1}, 0);
+                        layer.channel_mix_lerp_r = create_tensor(tn(LLM_TENSOR_CHANNEL_MIX_LERP_R, "weight", i), {n_embd, 1, 1}, 0);
+
+                        layer.channel_mix_key = create_tensor(tn(LLM_TENSOR_CHANNEL_MIX_KEY, "weight", i), {n_embd, ffn_size}, 0);
+                        layer.channel_mix_value = create_tensor(tn(LLM_TENSOR_CHANNEL_MIX_VALUE, "weight", i), {ffn_size, n_embd}, 0);
+                        layer.channel_mix_receptance = create_tensor(tn(LLM_TENSOR_CHANNEL_MIX_RECEPTANCE, "weight", i), {n_embd, n_embd}, 0);
+                    }
+
+                } break;
+            case LLM_ARCH_RWKV6QWEN2:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, 0);
+
+                    const int time_mix_extra_dim = hparams.time_mix_extra_dim;
+                    const int time_decay_extra_dim = hparams.time_decay_extra_dim;
+                    const int head_size = hparams.wkv_head_size;
+                    const int attn_hidden_size = n_embd;
+                    const int n_head_kv = hparams.n_head_kv();
+                    int attn_key_value_size;
+                    if (n_head_kv == 0 || attn_hidden_size / head_size == n_head_kv) {
+                        attn_key_value_size = attn_hidden_size;
+                    } else {
+                        attn_key_value_size = n_head_kv * head_size;
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.time_mix_w1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_W1, "weight", i), {n_embd, time_mix_extra_dim * 5}, 0);
+                        layer.time_mix_w2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_W2, "weight", i), {time_mix_extra_dim, n_embd, 5}, 0);
+
+                        layer.time_mix_lerp_x = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_X, "weight", i), {n_embd, 1, 1}, 0);
+                        layer.time_mix_lerp_fused = create_tensor(tn(LLM_TENSOR_TIME_MIX_LERP_FUSED, "weight", i), {n_embd, 1, 1, 5}, 0);
+
+                        layer.time_mix_first = create_tensor(tn(LLM_TENSOR_TIME_MIX_FIRST, "weight", i), {head_size, n_embd / head_size}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.time_mix_decay = create_tensor(tn(LLM_TENSOR_TIME_MIX_DECAY, "weight", i), {n_embd}, 0);
+                        layer.time_mix_decay_w1 = create_tensor(tn(LLM_TENSOR_TIME_MIX_DECAY_W1, "weight", i), {n_embd, time_decay_extra_dim}, 0);
+                        layer.time_mix_decay_w2 = create_tensor(tn(LLM_TENSOR_TIME_MIX_DECAY_W2, "weight", i), {time_decay_extra_dim, attn_hidden_size}, 0);
+                        layer.time_mix_key = create_tensor(tn(LLM_TENSOR_TIME_MIX_KEY, "weight", i), {n_embd, attn_key_value_size}, 0);
+                        layer.time_mix_value = create_tensor(tn(LLM_TENSOR_TIME_MIX_VALUE, "weight", i), {n_embd, attn_key_value_size}, 0);
+                        layer.time_mix_receptance = create_tensor(tn(LLM_TENSOR_TIME_MIX_RECEPTANCE, "weight", i), {attn_hidden_size, n_embd}, 0);
+                        layer.time_mix_gate = create_tensor(tn(LLM_TENSOR_TIME_MIX_GATE, "weight", i), {attn_hidden_size, n_embd}, 0);
+                        // optional bias tensors
+                        layer.time_mix_key_b = create_tensor(tn(LLM_TENSOR_TIME_MIX_KEY, "bias", i), {attn_key_value_size}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.time_mix_value_b = create_tensor(tn(LLM_TENSOR_TIME_MIX_VALUE, "bias", i), {attn_key_value_size}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        layer.time_mix_receptance_b = create_tensor(tn(LLM_TENSOR_TIME_MIX_RECEPTANCE, "bias", i), {attn_hidden_size}, llama_model_loader::TENSOR_NOT_REQUIRED);
+
+                        layer.time_mix_output = create_tensor(tn(LLM_TENSOR_TIME_MIX_OUTPUT, "weight", i), {n_embd, attn_hidden_size}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_CHAMELEON:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0);
+                        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k, n_head}, 0);
+                        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k, n_head_kv}, 0);
+                        layer.attn_q_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "bias", i),  {n_embd_head_k, n_head}, TENSOR_NOT_REQUIRED);
+                        layer.attn_k_norm_b = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "bias", i),  {n_embd_head_k, n_head_kv}, TENSOR_NOT_REQUIRED);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, 0);
+
+                        layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff}, 0);
+                        layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd}, 0);
+                        layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
+                    }
+                } break;
+            case LLM_ARCH_WAVTOKENIZER_DEC:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {hparams.n_embd_features, n_vocab}, 0);
+
+                    conv1d   = create_tensor(tn(LLM_TENSOR_CONV1D, "weight"), {7, hparams.n_embd_features, hparams.posnet.n_embd}, 0);
+                    conv1d_b = create_tensor(tn(LLM_TENSOR_CONV1D, "bias"),   {1, hparams.posnet.n_embd}, 0);
+
+                    // posnet
+                    {
+                        const int64_t n_embd = hparams.posnet.n_embd;
+
+                        for (uint32_t i = 0; i < hparams.posnet.n_layer; ++i) {
+                            auto & layer = layers[i].posnet;
+
+                            // posnet:
+                            //
+                            //  - resnet
+                            //  - resnet
+                            //  - attn
+                            //  - resnet
+                            //  - resnet
+                            //  - norm
+                            //
+                            switch (i) {
+                                case 0:
+                                case 1:
+                                case 3:
+                                case 4:
+                                    {
+                                        layer.norm1   = create_tensor(tn(LLM_TENSOR_POS_NET_NORM1, "weight", i), {1, n_embd}, 0);
+                                        layer.norm1_b = create_tensor(tn(LLM_TENSOR_POS_NET_NORM1, "bias",   i), {1, n_embd}, 0);
+
+                                        layer.conv1   = create_tensor(tn(LLM_TENSOR_POS_NET_CONV1, "weight", i), {3, n_embd, n_embd}, 0);
+                                        layer.conv1_b = create_tensor(tn(LLM_TENSOR_POS_NET_CONV1, "bias",   i), {1, n_embd}, 0);
+
+                                        layer.norm2   = create_tensor(tn(LLM_TENSOR_POS_NET_NORM2, "weight", i), {1, n_embd}, 0);
+                                        layer.norm2_b = create_tensor(tn(LLM_TENSOR_POS_NET_NORM2, "bias",   i), {1, n_embd}, 0);
+
+                                        layer.conv2   = create_tensor(tn(LLM_TENSOR_POS_NET_CONV2, "weight", i), {3, n_embd, n_embd}, 0);
+                                        layer.conv2_b = create_tensor(tn(LLM_TENSOR_POS_NET_CONV2, "bias",   i), {1, n_embd}, 0);
+                                    } break;
+                                case 2:
+                                    {
+                                        layer.attn_norm   = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_NORM, "weight", i), {1, n_embd}, 0);
+                                        layer.attn_norm_b = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_NORM, "bias",   i), {1, n_embd}, 0);
+
+                                        layer.attn_q      = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_Q,    "weight", i), {1, n_embd, n_embd}, 0);
+                                        layer.attn_q_b    = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_Q,    "bias",   i), {1, n_embd}, 0);
+
+                                        layer.attn_k      = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_K,    "weight", i), {1, n_embd, n_embd}, 0);
+                                        layer.attn_k_b    = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_K,    "bias",   i), {1, n_embd}, 0);
+
+                                        layer.attn_v      = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_V,    "weight", i), {1, n_embd, n_embd}, 0);
+                                        layer.attn_v_b    = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_V,    "bias",   i), {1, n_embd}, 0);
+
+                                        layer.attn_o      = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_OUT,  "weight", i), {1, n_embd, n_embd}, 0);
+                                        layer.attn_o_b    = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_OUT,  "bias",   i), {1, n_embd}, 0);
+                                    } break;
+                                case 5:
+                                    {
+                                        layer.norm   = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_NORM, "weight", i), {1, n_embd}, 0);
+                                        layer.norm_b = create_tensor(tn(LLM_TENSOR_POS_NET_ATTN_NORM, "bias",   i), {1, n_embd}, 0);
+                                    } break;
+                                default: GGML_ABORT("unknown posnet layer");
+                            };
+                        }
+                    }
+
+                    GGML_ASSERT(hparams.posnet.n_embd == hparams.convnext.n_embd);
+
+                    tok_norm   = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {hparams.posnet.n_embd}, 0);
+                    tok_norm_b = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"),   {hparams.posnet.n_embd}, 0);
+
+                    // convnext
+                    {
+                        const int64_t n_embd = hparams.convnext.n_embd;
+
+                        for (uint32_t i = 0; i < hparams.convnext.n_layer; ++i) {
+                            auto & layer = layers[i].convnext;
+
+                            layer.dw     = create_tensor(tn(LLM_TENSOR_CONVNEXT_DW,    "weight", i), {7, 1, n_embd}, 0);
+                            layer.dw_b   = create_tensor(tn(LLM_TENSOR_CONVNEXT_DW,    "bias",   i), {1, n_embd}, 0);
+
+                            layer.norm   = create_tensor(tn(LLM_TENSOR_CONVNEXT_NORM,  "weight", i), {n_embd}, 0);
+                            layer.norm_b = create_tensor(tn(LLM_TENSOR_CONVNEXT_NORM,  "bias",   i), {n_embd}, 0);
+
+                            layer.pw1    = create_tensor(tn(LLM_TENSOR_CONVNEXT_PW1,   "weight", i), {n_embd, n_ff}, 0);
+                            layer.pw1_b  = create_tensor(tn(LLM_TENSOR_CONVNEXT_PW1,   "bias",   i), {n_ff}, 0);
+
+                            layer.pw2    = create_tensor(tn(LLM_TENSOR_CONVNEXT_PW2,   "weight", i), {n_ff, n_embd}, 0);
+                            layer.pw2_b  = create_tensor(tn(LLM_TENSOR_CONVNEXT_PW2,   "bias",   i), {n_embd}, 0);
+
+                            layer.gamma  = create_tensor(tn(LLM_TENSOR_CONVNEXT_GAMMA, "weight", i), {n_embd}, 0);
+                        }
+
+                        // output
+                        output_norm   = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                        output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
+                    }
+
+                    output   = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {hparams.convnext.n_embd, n_embd}, 0);
+                    output_b = create_tensor(tn(LLM_TENSOR_OUTPUT, "bias"),   {n_embd}, 0);
+                } break;
+            default:
+                throw std::runtime_error("unknown architecture");
+        }
+
+        if (n_moved_tensors > 0) {
+            LLAMA_LOG_DEBUG("%s: tensor '%s' (%s) (and %d others) cannot be used with preferred buffer type %s, using %s instead\n",
+                __func__, first_moved_tensor->name, ggml_type_name(first_moved_tensor->type), n_moved_tensors - 1,
+                ggml_backend_buft_name(first_moved_from_buft), ggml_backend_buft_name(first_moved_to_buft));
+        }
+    }
+
+    ml.done_getting_tensors();
+
+    ml.init_mappings(true, use_mlock ? &pimpl->mlock_mmaps : nullptr);
+    pimpl->mappings.reserve(ml.mappings.size());
+
+    // create the backend buffers
+    std::vector> ctx_bufs;
+    ctx_bufs.reserve(ctx_map.size());
+
+    // Ensure we have enough capacity for the maximum backend buffer we will potentially create
+    const size_t n_max_backend_buffer = ctx_map.size() * ml.files.size();
+    pimpl->bufs.reserve(n_max_backend_buffer);
+
+    for (auto & it : ctx_map) {
+        ggml_backend_buffer_type_t buft = it.first;
+        ggml_context * ctx              = it.second;
+
+        // skip contexts without tensors
+        if (ggml_get_first_tensor(ctx) == nullptr) {
+            continue;
+        }
+
+        llama_buf_map buf_map;
+        buf_map.reserve(n_max_backend_buffer);
+
+        // check if it is possible to use buffer_from_host_ptr with this buffer type
+        ggml_backend_dev_t dev = ggml_backend_buft_get_device(buft);
+        if (!dev) {
+            // FIXME: workaround for CPU backend buft having a NULL device
+            dev = ggml_backend_dev_by_type(GGML_BACKEND_DEVICE_TYPE_CPU);
+        }
+        ggml_backend_dev_props props;
+        ggml_backend_dev_get_props(dev, &props);
+        bool buffer_from_host_ptr_supported = props.caps.buffer_from_host_ptr;
+        bool is_default_buft = buft == ggml_backend_dev_buffer_type(dev);
+
+        if (ml.use_mmap && use_mmap_buffer && buffer_from_host_ptr_supported && is_default_buft) {
+            for (uint32_t idx = 0; idx < ml.files.size(); idx++) {
+                // only the mmap region containing the tensors in the model is mapped to the backend buffer
+                // this is important for metal with apple silicon: if the entire model could be mapped to a metal buffer, then we could just use metal for all layers
+                // this allows using partial offloading when the model size exceeds the metal buffer size, but not the RAM size
+                void * addr = nullptr;
+                size_t first, last; // NOLINT
+                ml.get_mapping_range(&first, &last, &addr, idx, ctx);
+                if (first >= last) {
+                    continue;
+                }
+                const size_t max_size = ggml_get_max_tensor_size(ctx);
+                ggml_backend_buffer_t buf = ggml_backend_dev_buffer_from_host_ptr(dev, (char *) addr + first, last - first, max_size);
+                if (buf == nullptr) {
+                    throw std::runtime_error(format("unable to allocate %s buffer", ggml_backend_buft_name(buft)));
+                }
+                pimpl->bufs.emplace_back(buf);
+                buf_map.emplace(idx, buf);
+            }
+        }
+        else {
+            ggml_backend_buffer_t buf = ggml_backend_alloc_ctx_tensors_from_buft(ctx, buft);
+            if (buf == nullptr) {
+                throw std::runtime_error(format("unable to allocate %s buffer", ggml_backend_buft_name(buft)));
+            }
+            pimpl->bufs.emplace_back(buf);
+            if (use_mlock && ggml_backend_buffer_is_host(buf)) {
+                pimpl->mlock_bufs.emplace_back(new llama_mlock);
+                auto & mlock_buf = pimpl->mlock_bufs.back();
+                mlock_buf->init   (ggml_backend_buffer_get_base(buf));
+                mlock_buf->grow_to(ggml_backend_buffer_get_size(buf));
+            }
+            for (uint32_t idx = 0; idx < ml.files.size(); idx++) {
+                buf_map.emplace(idx, buf);
+            }
+        }
+
+        if (pimpl->bufs.empty()) {
+            throw std::runtime_error("failed to allocate buffer");
+        }
+
+        for (auto & buf : buf_map) {
+            // indicate that this buffer contains weights
+            // this is used by ggml_backend_sched to improve op scheduling: ops that use a weight are preferably scheduled to the backend that contains the weight
+            ggml_backend_buffer_set_usage(buf.second, GGML_BACKEND_BUFFER_USAGE_WEIGHTS);
+        }
+
+        ctx_bufs.emplace_back(ctx, buf_map);
+    }
+
+    if (llama_supports_gpu_offload()) {
+        const int n_gpu = std::min(n_gpu_layers, int(hparams.n_layer));
+
+        LLAMA_LOG_INFO("%s: offloading %d repeating layers to GPU\n", __func__, n_gpu);
+        if (n_gpu_layers > (int) hparams.n_layer) {
+            LLAMA_LOG_INFO("%s: offloading output layer to GPU\n", __func__);
+        }
+
+        const int max_backend_supported_layers = hparams.n_layer + 1;
+        const int max_offloadable_layers       = hparams.n_layer + 1;
+
+        LLAMA_LOG_INFO("%s: offloaded %d/%d layers to GPU\n", __func__, std::min(n_gpu_layers, max_offloadable_layers), max_backend_supported_layers);
+    }
+
+    // print memory requirements per buffer type
+    for (auto & buf : pimpl->bufs) {
+        LLAMA_LOG_INFO("%s: %12s model buffer size = %8.2f MiB\n", __func__, ggml_backend_buffer_name(buf.get()), ggml_backend_buffer_get_size(buf.get()) / 1024.0 / 1024.0);
+    }
+
+    // populate tensors_by_name
+    for (auto & ctx : pimpl->ctxs) {
+        for (auto * cur = ggml_get_first_tensor(ctx.get()); cur != NULL; cur = ggml_get_next_tensor(ctx.get(), cur)) {
+            tensors_by_name.emplace_back(ggml_get_name(cur), cur);
+        }
+    }
+
+    // load tensor data
+    for (auto & it : ctx_bufs) {
+        ggml_context * ctx = it.first;
+        auto & bufs = it.second;
+        if (!ml.load_all_data(ctx, bufs, use_mlock ? &pimpl->mlock_mmaps : NULL, params.progress_callback, params.progress_callback_user_data)) {
+            return false;
+        }
+    }
+
+    if (use_mmap_buffer) {
+        for (auto & mapping : ml.mappings) {
+            pimpl->mappings.emplace_back(std::move(mapping));
+        }
+    }
+
+    return true;
+}
+
+std::string llama_model::arch_name() const {
+    return llm_arch_name(arch);
+}
+
+std::string llama_model::type_name() const {
+    return llm_type_name(type);
+}
+
+std::string llama_model::desc() const {
+    return pimpl->desc_str;
+}
+
+size_t llama_model::size() const {
+    return pimpl->n_bytes;
+}
+
+size_t llama_model::max_nodes() const {
+    return std::max(8192, tensors_by_name.size()*5);
+}
+
+size_t llama_model::n_devices() const {
+    return devices.size();
+}
+
+uint64_t llama_model::n_elements() const {
+    return pimpl->n_elements;
+}
+
+void llama_model::print_info() const {
+    const char * rope_scaling_type = LLAMA_ROPE_SCALING_TYPES.at(hparams.rope_scaling_type_train);
+
+    auto print_f = [](const std::function & f, uint32_t n) {
+        bool is_var = false;
+
+        std::vector v;
+        for (uint32_t i = 0; i < n; ++i) {
+            v.push_back(f(i));
+            if (v[i] != v[0]) {
+                is_var = true;
+            }
+        }
+
+        std::stringstream ss;
+
+        if (is_var) {
+            ss << "[";
+            for (uint32_t i = 0; i < n; ++i) {
+                ss << v[i];
+                if (i < n - 1) {
+                    ss << ", ";
                 }
             }
             ss << "]";
@@ -1859,11 +3550,7 @@ void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
     };
 
     // hparams
-    LLAMA_LOG_INFO("%s: format           = %s\n",     __func__, llama_file_version_name(ml.fver));
-    LLAMA_LOG_INFO("%s: arch             = %s\n",     __func__, llm_arch_name(model.arch));
-    LLAMA_LOG_INFO("%s: vocab type       = %s\n",     __func__, llama_model_vocab_type_name(vocab.type));
-    LLAMA_LOG_INFO("%s: n_vocab          = %u\n",     __func__, hparams.n_vocab);
-    LLAMA_LOG_INFO("%s: n_merges         = %u\n",     __func__, (int) vocab.bpe_ranks.size());
+    LLAMA_LOG_INFO("%s: arch             = %s\n",     __func__, arch_name().c_str());
     LLAMA_LOG_INFO("%s: vocab_only       = %d\n",     __func__, hparams.vocab_only);
 
     if (!hparams.vocab_only) {
@@ -1902,60 +3589,28 @@ void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
         LLAMA_LOG_INFO("%s: ssm_dt_b_c_rms   = %d\n",     __func__, hparams.ssm_dt_b_c_rms);
     }
 
-    LLAMA_LOG_INFO("%s: model type       = %s\n",     __func__, llama_model_type_name(model).c_str());
-    LLAMA_LOG_INFO("%s: model ftype      = %s\n",     __func__, llama_model_ftype_name(model).c_str());
-    if (ml.n_elements >= 1e12) {
-        LLAMA_LOG_INFO("%s: model params     = %.2f T\n", __func__, ml.n_elements*1e-12);
-    } else if (ml.n_elements >= 1e9) {
-        LLAMA_LOG_INFO("%s: model params     = %.2f B\n", __func__, ml.n_elements*1e-9);
-    } else if (ml.n_elements >= 1e6) {
-        LLAMA_LOG_INFO("%s: model params     = %.2f M\n", __func__, ml.n_elements*1e-6);
+    LLAMA_LOG_INFO("%s: model type       = %s\n",     __func__, type_name().c_str());
+    if (pimpl->n_elements >= 1e12) {
+        LLAMA_LOG_INFO("%s: model params     = %.2f T\n", __func__, pimpl->n_elements*1e-12);
+    } else if (pimpl->n_elements >= 1e9) {
+        LLAMA_LOG_INFO("%s: model params     = %.2f B\n", __func__, pimpl->n_elements*1e-9);
+    } else if (pimpl->n_elements >= 1e6) {
+        LLAMA_LOG_INFO("%s: model params     = %.2f M\n", __func__, pimpl->n_elements*1e-6);
     } else {
-        LLAMA_LOG_INFO("%s: model params     = %.2f K\n", __func__, ml.n_elements*1e-3);
-    }
-    if (ml.n_bytes < GiB) {
-        LLAMA_LOG_INFO("%s: model size       = %.2f MiB (%.2f BPW) \n", __func__, ml.n_bytes/1024.0/1024.0,        ml.n_bytes*8.0/ml.n_elements);
-    } else {
-        LLAMA_LOG_INFO("%s: model size       = %.2f GiB (%.2f BPW) \n", __func__, ml.n_bytes/1024.0/1024.0/1024.0, ml.n_bytes*8.0/ml.n_elements);
+        LLAMA_LOG_INFO("%s: model params     = %.2f K\n", __func__, pimpl->n_elements*1e-3);
     }
 
     // general kv
-    LLAMA_LOG_INFO("%s: general.name     = %s\n",    __func__, model.name.c_str());
-
-    // special tokens
-    if (vocab.special_bos_id  != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: BOS token        = %d '%s'\n", __func__, vocab.special_bos_id,     vocab.id_to_token[vocab.special_bos_id].text.c_str() );  }
-    if (vocab.special_eos_id  != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: EOS token        = %d '%s'\n", __func__, vocab.special_eos_id,     vocab.id_to_token[vocab.special_eos_id].text.c_str() );  }
-    if (vocab.special_eot_id  != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: EOT token        = %d '%s'\n", __func__, vocab.special_eot_id,     vocab.id_to_token[vocab.special_eot_id].text.c_str() );  }
-    if (vocab.special_eom_id  != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: EOM token        = %d '%s'\n", __func__, vocab.special_eom_id,     vocab.id_to_token[vocab.special_eom_id].text.c_str() );  }
-    if (vocab.special_unk_id  != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: UNK token        = %d '%s'\n", __func__, vocab.special_unk_id,     vocab.id_to_token[vocab.special_unk_id].text.c_str() );  }
-    if (vocab.special_sep_id  != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: SEP token        = %d '%s'\n", __func__, vocab.special_sep_id,     vocab.id_to_token[vocab.special_sep_id].text.c_str() );  }
-    if (vocab.special_pad_id  != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: PAD token        = %d '%s'\n", __func__, vocab.special_pad_id,     vocab.id_to_token[vocab.special_pad_id].text.c_str() );  }
-    if (vocab.special_cls_id  != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: CLS token        = %d '%s'\n", __func__, vocab.special_cls_id,     vocab.id_to_token[vocab.special_cls_id].text.c_str() );  }
-    if (vocab.special_mask_id != LLAMA_TOKEN_NULL)    { LLAMA_LOG_INFO( "%s: MASK token       = %d '%s'\n", __func__, vocab.special_mask_id,    vocab.id_to_token[vocab.special_mask_id].text.c_str() ); }
-
-    if (vocab.linefeed_id != LLAMA_TOKEN_NULL)        { LLAMA_LOG_INFO( "%s: LF token         = %d '%s'\n", __func__, vocab.linefeed_id,        vocab.id_to_token[vocab.linefeed_id].text.c_str() ); }
-
-    if (vocab.special_fim_pre_id != LLAMA_TOKEN_NULL) { LLAMA_LOG_INFO( "%s: FIM PRE token    = %d '%s'\n", __func__, vocab.special_fim_pre_id, vocab.id_to_token[vocab.special_fim_pre_id].text.c_str() ); }
-    if (vocab.special_fim_suf_id != LLAMA_TOKEN_NULL) { LLAMA_LOG_INFO( "%s: FIM SUF token    = %d '%s'\n", __func__, vocab.special_fim_suf_id, vocab.id_to_token[vocab.special_fim_suf_id].text.c_str() ); }
-    if (vocab.special_fim_mid_id != LLAMA_TOKEN_NULL) { LLAMA_LOG_INFO( "%s: FIM MID token    = %d '%s'\n", __func__, vocab.special_fim_mid_id, vocab.id_to_token[vocab.special_fim_mid_id].text.c_str() ); }
-    if (vocab.special_fim_pad_id != LLAMA_TOKEN_NULL) { LLAMA_LOG_INFO( "%s: FIM PAD token    = %d '%s'\n", __func__, vocab.special_fim_pad_id, vocab.id_to_token[vocab.special_fim_pad_id].text.c_str() ); }
-    if (vocab.special_fim_rep_id != LLAMA_TOKEN_NULL) { LLAMA_LOG_INFO( "%s: FIM REP token    = %d '%s'\n", __func__, vocab.special_fim_rep_id, vocab.id_to_token[vocab.special_fim_rep_id].text.c_str() ); }
-    if (vocab.special_fim_sep_id != LLAMA_TOKEN_NULL) { LLAMA_LOG_INFO( "%s: FIM SEP token    = %d '%s'\n", __func__, vocab.special_fim_sep_id, vocab.id_to_token[vocab.special_fim_sep_id].text.c_str() ); }
-
-    for (const auto & id : vocab.special_eog_ids) {
-        LLAMA_LOG_INFO( "%s: EOG token        = %d '%s'\n", __func__, id, vocab.id_to_token[id].text.c_str() );
-    }
-
-    LLAMA_LOG_INFO("%s: max token length = %d\n", __func__, vocab.max_token_len);
+    LLAMA_LOG_INFO("%s: general.name     = %s\n",    __func__, name.c_str());
 
-    if (model.arch == LLM_ARCH_DEEPSEEK) {
+    if (arch == LLM_ARCH_DEEPSEEK) {
         LLAMA_LOG_INFO("%s: n_layer_dense_lead   = %d\n",     __func__, hparams.n_layer_dense_lead);
         LLAMA_LOG_INFO("%s: n_ff_exp             = %d\n",     __func__, hparams.n_ff_exp);
         LLAMA_LOG_INFO("%s: n_expert_shared      = %d\n",     __func__, hparams.n_expert_shared);
         LLAMA_LOG_INFO("%s: expert_weights_scale = %.1f\n",   __func__, hparams.expert_weights_scale);
     }
 
-    if (model.arch == LLM_ARCH_DEEPSEEK2) {
+    if (arch == LLM_ARCH_DEEPSEEK2) {
         LLAMA_LOG_INFO("%s: n_layer_dense_lead   = %d\n",     __func__, hparams.n_layer_dense_lead);
         LLAMA_LOG_INFO("%s: n_lora_q             = %d\n",     __func__, hparams.n_lora_q);
         LLAMA_LOG_INFO("%s: n_lora_kv            = %d\n",     __func__, hparams.n_lora_kv);
@@ -1967,16 +3622,88 @@ void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
         LLAMA_LOG_INFO("%s: rope_yarn_log_mul    = %.4f\n",   __func__, hparams.rope_yarn_log_mul);
     }
 
-    if (model.arch == LLM_ARCH_QWEN2MOE) {
+    if (arch == LLM_ARCH_QWEN2MOE) {
         LLAMA_LOG_INFO("%s: n_ff_exp         = %d\n",     __func__, hparams.n_ff_exp);
         LLAMA_LOG_INFO("%s: n_ff_shexp       = %d\n",     __func__, hparams.n_ff_shexp);
     }
 
-    if (model.arch == LLM_ARCH_MINICPM || model.arch == LLM_ARCH_GRANITE || model.arch == LLM_ARCH_GRANITE_MOE) {
+    if (arch == LLM_ARCH_MINICPM || arch == LLM_ARCH_GRANITE || arch == LLM_ARCH_GRANITE_MOE) {
         LLAMA_LOG_INFO("%s: f_embedding_scale = %f\n", __func__, hparams.f_embedding_scale);
         LLAMA_LOG_INFO("%s: f_residual_scale  = %f\n", __func__, hparams.f_residual_scale);
         LLAMA_LOG_INFO("%s: f_attention_scale = %f\n", __func__, hparams.f_attention_scale);
     }
+
+    vocab.print_info();
+}
+
+ggml_backend_dev_t llama_model::dev_layer(int il) const {
+    return pimpl->dev_layer.at(il).dev;
+}
+
+ggml_backend_dev_t llama_model::dev_output() const {
+    return pimpl->dev_output.dev;
+}
+
+template
+static bool buft_supported(ggml_backend_buffer_type_t buft, ggml_backend_dev_t dev, F & fn) {
+    ggml_init_params params = {
+        /*.mem_size   =*/ ggml_tensor_overhead()*8,
+        /*.mem_buffer =*/ NULL,
+        /*.no_alloc   =*/ true,
+    };
+
+    ggml_context_ptr ctx { ggml_init(params) };
+    if (!ctx) {
+        throw std::runtime_error(format("failed to create ggml context"));
+    }
+
+    ggml_backend_buffer_ptr buf { ggml_backend_buft_alloc_buffer(buft, 0) };
+    ggml_tensor * op_tensor = fn(ctx.get());
+    for (int i = 0; i < GGML_MAX_SRC; i++) {
+        if (op_tensor->src[i] != nullptr) {
+            assert(op_tensor->src[i]->buffer == nullptr);
+            op_tensor->src[i]->buffer = buf.get();
+        }
+    }
+
+    bool op_supported = ggml_backend_dev_supports_op(dev, op_tensor);
+
+    return op_supported;
+}
+
+template
+static ggml_backend_buffer_type_t select_buft(const buft_list_t & buft_list, const F & fn) {
+    for (const auto & cur : buft_list) {
+        ggml_backend_dev_t cur_dev = cur.first;
+        ggml_backend_buffer_type_t cur_buft = cur.second;
+        if (buft_supported(cur_buft, cur_dev, fn)) {
+            return cur_buft;
+        }
+    }
+
+    throw std::runtime_error(format("no suitable buffer type found"));
+}
+
+ggml_backend_buffer_type_t llama_model::select_buft(int il) const {
+    return ::select_buft(
+            *pimpl->dev_layer.at(il).buft_list,
+            [&](ggml_context * ctx) {
+                ggml_tensor * cur = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, hparams.n_embd);
+                ggml_tensor * layer_dir = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, hparams.n_embd);
+                return ggml_add(ctx, cur, layer_dir);
+            });
+}
+
+const struct ggml_tensor * llama_model::get_tensor(const char * name) const {
+    auto it = std::find_if(tensors_by_name.begin(), tensors_by_name.end(),
+            [name](const std::pair & it) {
+                return it.first == name;
+            });
+    if (it == tensors_by_name.end()) {
+        return nullptr;
+    }
+
+    return it->second;
 }
 
 //
@@ -2008,6 +3735,10 @@ struct llama_model_params llama_model_default_params() {
     return result;
 }
 
+const struct llama_vocab * llama_model_get_vocab(const struct llama_model * model) {
+    return &model->vocab;
+}
+
 void llama_free_model(struct llama_model * model) {
     llama_model_free(model);
 }
@@ -2016,31 +3747,43 @@ void llama_model_free(struct llama_model * model) {
     delete model;
 }
 
-enum llama_vocab_type llama_vocab_type(const struct llama_model * model) {
-    return model->vocab.type;
+int32_t llama_model_n_ctx_train(const struct llama_model * model) {
+    return model->hparams.n_ctx_train;
+}
+
+int32_t llama_model_n_embd(const struct llama_model * model) {
+    return model->hparams.n_embd;
+}
+
+int32_t llama_model_n_layer(const struct llama_model * model) {
+    return model->hparams.n_layer;
 }
 
-int32_t llama_n_vocab(const struct llama_model * model) {
-    return model->hparams.n_vocab;
+int32_t llama_model_n_head(const struct llama_model * model) {
+    return model->hparams.n_head();
 }
 
+// deprecated
 int32_t llama_n_ctx_train(const struct llama_model * model) {
-    return model->hparams.n_ctx_train;
+    return llama_model_n_ctx_train(model);
 }
 
+// deprecated
 int32_t llama_n_embd(const struct llama_model * model) {
-    return model->hparams.n_embd;
+    return llama_model_n_embd(model);
 }
 
+// deprecated
 int32_t llama_n_layer(const struct llama_model * model) {
-    return model->hparams.n_layer;
+    return llama_model_n_layer(model);
 }
 
+// deprecated
 int32_t llama_n_head(const struct llama_model * model) {
-    return model->hparams.n_head();
+    return llama_model_n_head(model);
 }
 
-enum llama_rope_type llama_rope_type(const struct llama_model * model) {
+enum llama_rope_type llama_model_rope_type(const struct llama_model * model) {
     switch (model->arch) {
         // these models do not use RoPE
         case LLM_ARCH_GPT2:
@@ -2054,6 +3797,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
         case LLM_ARCH_T5ENCODER:
         case LLM_ARCH_JAIS:
         case LLM_ARCH_RWKV6:
+        case LLM_ARCH_RWKV6QWEN2:
         case LLM_ARCH_WAVTOKENIZER_DEC:
             return LLAMA_ROPE_TYPE_NONE;
 
@@ -2094,6 +3838,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
         case LLM_ARCH_OLMOE:
         case LLM_ARCH_PHI2:
         case LLM_ARCH_PHI3:
+        case LLM_ARCH_PHIMOE:
         case LLM_ARCH_GEMMA:
         case LLM_ARCH_GEMMA2:
         case LLM_ARCH_STARCODER2:
@@ -2116,7 +3861,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
     return LLAMA_ROPE_TYPE_NONE;
 }
 
-float llama_rope_freq_scale_train(const struct llama_model * model) {
+float llama_model_rope_freq_scale_train(const struct llama_model * model) {
     return model->hparams.rope_freq_scale_train;
 }
 
@@ -2160,18 +3905,24 @@ int32_t llama_model_meta_val_str_by_index(const struct llama_model * model, int3
 }
 
 int32_t llama_model_desc(const struct llama_model * model, char * buf, size_t buf_size) {
-    return snprintf(buf, buf_size, "%s %s %s",
-            llama_model_arch_name (*model).c_str(),
-            llama_model_type_name (*model).c_str(),
-            llama_model_ftype_name(*model).c_str());
+    return snprintf(buf, buf_size, "%s", model->desc().c_str());
 }
 
 uint64_t llama_model_size(const struct llama_model * model) {
-    return model->n_bytes;
+    return model->size();
+}
+
+const char * llama_model_chat_template(const struct llama_model * model) {
+    const auto & it = model->gguf_kv.find(LLM_KV(model->arch)(LLM_KV_TOKENIZER_CHAT_TEMPLATE));
+    if (it == model->gguf_kv.end()) {
+        return nullptr;
+    }
+
+    return it->second.c_str();
 }
 
 uint64_t llama_model_n_params(const struct llama_model * model) {
-    return model->n_elements;
+    return model->n_elements();
 }
 
 bool llama_model_has_encoder(const struct llama_model * model) {
@@ -2197,6 +3948,7 @@ bool llama_model_is_recurrent(const struct llama_model * model) {
     switch (model->arch) {
         case LLM_ARCH_MAMBA:  return true;
         case LLM_ARCH_RWKV6:  return true;
+        case LLM_ARCH_RWKV6QWEN2: return true;
         default:              return false;
     }
 }