mnist: fix segmentation fault (ggml/1227)

llama/ggml: add LLM training support (llama/10544)

Add `--no-op-offload` to improve `-ot` pp perf in MoE models like llama4 400B (llama/13386)

CUDA: fix bad asserts for partial offload (llama/13337)

CUDA: fix logic for clearing padding with -ngl 0 (llama/13320)

CUDA: fix q_nope_absorbed prec for DS 2 Lite f16 (llama/13137)

ggml: move fp16/bf16 conversion optimizations to CPU backend + export conversion APIs (llama/13107)

rpc : do not wait for response when sending RPC_CMD_SET_TENSOR (llama/12943)

ggml : fix ggml_gallocr_ptr type (ggml/1205)

rpc : add RPC_CMD_HELLO (llama/12955)

ggml : Depthwise 2D convolution (ggml/1152)

ggml : add bilinear upscale support (ggml/1185)

ggml : add more generic custom op, remove deprecated custom ops (ggml/1183)

metal : improve FA + improve MoE (llama/12612)

rpc : send hash when tensor data is above some fixed threshold (llama/12496)

llama: Add support for RWKV v7 architecture (llama/12412)

ggml-cpu: Faster IQ1 mul_mat_vec on AVX2 using BMI2 instructions (llama/12154)

ggml : portability fixes for VS 2017 (llama/12150)

ggml : upgrade init_tensor API to return a ggml_status (llama/11854)

ggml-cpu: Support s390x SIMD Instruction Set (llama/12019)

ggml-cpu: Add CPU backend support for KleidiAI library (llama/11390)

repo : update links to new url (llama/11886)

cleanup: fix compile warnings associated with gnu_printf (llama/11811)

vulkan: Make Vulkan optional at runtime (ggml/11493). (llama/11494)

CUDA: use mma PTX instructions for FlashAttention (llama/11583)

rpc : early register backend devices (llama/11262)

CUDA: backwards pass for misc. ops, add tests (llama/11257)

RoPE: fix back, CUDA support for back + noncont. (llama/11240)

GGUF: C++ refactor, backend support, misc fixes (skip) (llama/11030)

llama: add support for QRWKV6 model architecture (llama/11001)

GGUF: C++ refactor, backend support, misc fixes (llama/11030)

tts : add OuteTTS support (llama/10784)

llama : add Qwen2VL support + multimodal RoPE (llama/10361)

Introducing experimental OpenCL backend with support for Qualcomm Adreno GPUs (llama/10693)

ggml: load all backends from a user-provided search path (llama/10699)

ggml : refactor online repacking (llama/10446)

ggml : remove old files (skip) (#0)

ggml : add `GGML_PAD_REFLECT_1D` operation (ggml/1034)

ggml-cpu: support IQ4_NL_4_4 by runtime repack (llama/10541)

ggml : add support for dynamic loading of backends (llama/10469)

ggml: new optimization interface (ggml/988)

backend cpu: add online flow for aarch64 Q4_0 GEMV/GEMM kernels (llama/9921)

ggml : build backends as libraries (llama/10256)

metal : optimize FA kernels (llama/10171)

Optimize RWKV6 Operator Naming and Implement Multi-core CPU/ SYCL Acceleration (llama/10133)

ggml : move CPU backend to a separate file (llama/10144)

llama : add simple-chat example (llama/10124)

llama : use smart pointers for ggml resources (llama/10117)

ggml : remove ggml_scratch (llama/10121)

kompute: add backend registry / device interfaces (llama/10045)