Duplicate from allura-forge/expr-rp-sft-mix

Co-authored-by: Fizz 🏳️‍⚧️ <Fizzarolli@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,59 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,21 @@
+---
+dataset_info:
+  features:
+  - name: conversations
+    list:
+    - name: from
+      dtype: string
+    - name: value
+      dtype: string
+  splits:
+  - name: train
+    num_bytes: 617948242
+    num_examples: 46586
+  download_size: 369997405
+  dataset_size: 617948242
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-*
+---

data/train-00000-of-00002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb058ca314289efee09d52abafe64bfbcd2b41e905e6b4aa9d26632e663e083d
+size 185764459

data/train-00001-of-00002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0245478c1157097b5041fffbc6cfe7914d9f5809d5478f211da514147b25ccc9
+size 184232946

main.py

@@ -0,0 +1,341 @@
+from datasets import load_dataset, concatenate_datasets, Features, Value
+from huggingface_hub import HfApi
+from functools import partial
+
+def convert_openai_to_sharegpt(sample):
+    convo = []
+    for msg in sample["conversations"]:
+        if msg["content"] == "":
+            continue
+        match msg["role"]:
+            case "user":
+                convo.append({
+                    "from": "human",
+                    "value": msg["content"]
+                })
+            case "assistant":
+                convo.append({
+                    "from": "gpt",
+                    "value": msg["content"]
+                })
+            case other:
+                convo.append({
+                    "from": other,
+                    "value": msg["content"]
+                })
+    return {"conversations": convo}
+
+def convert_openai_to_sharegpt_thonking_added(sample):
+    convo = [{"from": "system", "value": "When responding, think step by step in a set of XML tags (<think> and </think>) and consider your response deeply, then close the </think> tag and write the rest of your response normally."}]
+    for msg in sample["conversations"]:
+        if msg["content"] == "":
+            continue
+        match msg["role"]:
+            case "user":
+                convo.append({
+                    "from": "human",
+                    "value": msg["content"]
+                })
+            case "assistant":
+                convo.append({
+                    "from": "gpt",
+                    "value": msg["content"]
+                })
+            case other:
+                convo.append({
+                    "from": other,
+                    "value": msg["content"]
+                })
+    return {"conversations": convo}
+
+def convert_ruozhiba_to_sharegpt(sample):
+    return {
+        "conversations": [
+            {
+                "from": "human",
+                "value": sample["instruction"]
+            },
+            {
+                "from": "gpt",
+                "value": sample["output"]
+            }
+        ]
+    }
+
+def convert_zh_to_sharegpt(sample):
+    conversation = sample["conversations"]
+    conversation.insert(0, {
+        "from": "system",
+        "value": sample["system_prompt"]
+    })
+    return {
+        "conversations": conversation
+    }
+
+def convert_opencai(sample):
+    convo = []
+    for idx, msg in enumerate(sample["conversations"]):
+        convo.append({
+            "from": "human" if idx % 2 == 0 else "gpt",
+            "value": f"{msg['author']}: {msg['message']}"
+        })
+    return {"conversations": convo}
+
+def convert_opc_sft_stage1(sample):
+    return {
+        "conversations": [
+            {
+                "from": "human",
+                "value": sample["instruction"]
+            },
+            {
+                "from": "gpt",
+                "value": sample["output"]
+            }
+        ]
+    }
+
+def convert_original_s1k(sample, think_key, attempt_key):
+    return {
+        "conversations": [
+            {"from": "system", "value": "When responding, think step by step in a set of XML tags (<think> and </think>) and consider your response deeply, then close the </think> tag and write the rest of your response normally."},
+            {
+                "from": "human",
+                "value": sample["question"]
+            },
+            {
+                "from": "gpt",
+                "value": f"""<think>\n{sample[think_key]}\n</think>\n{sample[attempt_key]}"""
+            }
+        ]
+    }
+
+def convert_freedom_medical(sample):
+    return {
+        "conversations": [
+            {"from": "system", "value": "When responding, think step by step in a set of XML tags (<think> and </think>) and consider your response deeply, then close the </think> tag and write the rest of your response normally."},
+            {
+                "from": "human",
+                "value": sample["question"]
+            },
+            {
+                "from": "gpt",
+                "value": f"""<think>\n{sample["reasoning (reasoning_content)"]}\n</think>\n{sample["response (content)"]}"""
+            }
+        ]
+    }
+
+def convert_chinese_deepseek_r1_distill_data(sample):
+    return {
+        "conversations": [
+            {
+                "from": "human",
+                "value": sample["input"]
+            },
+            {
+                "from": "gpt",
+                "value": f"""{sample['content'].strip()}"""
+            }
+        ]
+    }
+
+def filter_synthkink(sample):
+    # Make a copy of the story and check if it contains <think> tags
+    story = sample["story"].strip()
+    
+    # If the story contains <think> tags, strip them out
+    if "<think>" in story and "</think>" in story:
+        # Find the positions of the tags
+        think_start = story.find("<think>")
+        think_end = story.find("</think>") + len("</think>")
+        
+        # Remove the tags and everything between them
+        final_response = story[:think_start] + story[think_end:]
+        
+        # Clean up any extra newlines that might be left
+        final_response = final_response.strip()
+    else:
+        # If no think tags, use the original story
+        final_response = story
+    
+    # Check if the final response is longer than 500 characters
+    return len(final_response) > 500
+
+def convert_synthkink(sample):
+    story = sample["story"].strip()
+    # If the story contains <think> tags, strip them out
+    if "<think>" in story and "</think>" in story:
+        # Find the positions of the tags
+        think_start = story.find("<think>")
+        think_end = story.find("</think>") + len("</think>")
+        
+        # Remove the tags and everything between them
+        final_response = story[:think_start] + story[think_end:]
+        
+        # Clean up any extra newlines that might be left
+        final_response = final_response.strip()
+    else:
+        # If no think tags, use the original story
+        final_response = story
+    
+    return {
+        "conversations": [
+            {
+                "from": "human",
+                "value": sample["prompt"]
+            },
+            {
+                "from": "gpt",
+                "value": final_response
+            }
+        ]
+    }
+
+def convert_helpsteer(sample):
+    context = [{"from": "human" if msg["role"] == "user" else "gpt", "value": msg["content"]} for msg in sample["context"]]
+    if sample["overall_preference"] > 0:
+        return {
+            "conversations": context + [
+                {
+                    "from": "gpt",
+                    "value": sample["response2"]
+                }
+            ]
+        }
+    else:
+        return {
+            "conversations": context + [
+                {
+                    "from": "gpt",
+                    "value": sample["response1"]
+                }
+            ]
+        }
+
+datasets = [
+    # general
+    (
+        load_dataset("QuixiAI/SystemChat-2.0", split="train", data_files={"train": "SystemChat_filtered.jsonl"})
+            .rename_column("messages", "conversations")
+            .map(convert_openai_to_sharegpt), 
+        1.0
+    ),
+
+    # claude
+    (
+        load_dataset("anthracite-org/kalo-opus-instruct-22k-no-refusal", split="train").map(lambda x: {"conversations": [{"from": "system", "value": "You are Luna, a helpful and harmless language model by Allura."}] + x["conversations"][1:]}).map(lambda x: {"conversations": [{"from": msg["from"], "value": msg["value"].replace("Claude", "Luna").replace("claude", "luna")} for msg in x["conversations"]]}),
+        None
+    ),
+
+
+    # rp
+    (
+        load_dataset("Delta-Vector/Orion-OpenCAI-ShareGPT", split="train"),
+        None
+    ),
+    (
+        load_dataset("Delta-Vector/Orion-PIPPA-Cleaned-V2", split="train"),
+        None
+    ),
+    (
+        load_dataset("grimulkan/LimaRP-augmented", split="train").remove_columns(["id"]),
+        None
+    ),
+    (
+        load_dataset("Delta-Vector/Orion-PIPPA-Cleaned-V2", split="train"),
+        None
+    ),
+    (
+        load_dataset("grimulkan/LimaRP-augmented", split="train").remove_columns(["id"]),
+        None
+    ),
+    (
+        load_dataset("Delta-Vector/Orion-Personamaxx-Logs-21K", split="train").map(lambda x: {"conversations": [{"from":msg["from"],"value":msg["value"]} for msg in x["conversations"]]}, features=Features({
+    "conversations": [{"from": Value("string"), "value": Value("string")}]
+    # add other columns you want to keep
+})),
+        1.0
+    ),
+
+    # story
+    (
+        load_dataset("Delta-Vector/Orion-Fiction-Writer", split="train"),
+        None
+    ),
+    (
+        load_dataset("Delta-Vector/Orion-Praxis-Co-Writer", split="train"),
+        None
+    ),
+
+    # beep boop
+    (
+        load_dataset("allura-forge/luna-hardcoded-expr", split="train")
+            .rename_column("messages", "conversations")
+            .map(convert_openai_to_sharegpt), 
+        None
+    ),
+    (
+        load_dataset("allura-forge/luna-hardcoded-expr", split="train")
+            .rename_column("messages", "conversations")
+            .map(convert_openai_to_sharegpt), 
+        None
+    ),
+    (
+        load_dataset("allura-forge/luna-hardcoded-expr", split="train")
+            .rename_column("messages", "conversations")
+            .map(convert_openai_to_sharegpt), 
+        None
+    ),
+    (
+        load_dataset("allura-forge/luna-hardcoded-expr", split="train")
+            .rename_column("messages", "conversations")
+            .map(convert_openai_to_sharegpt), 
+        None
+    ),
+    (
+        load_dataset("allura-forge/luna-hardcoded-expr", split="train")
+            .rename_column("messages", "conversations")
+            .map(convert_openai_to_sharegpt), 
+        None
+    ),
+
+]
+
+# Get the minimum number of entries across all datasets, and truncate all datasets to this length so we don't overrepresent any dataset
+#max_entries = float('inf')
+#for dataset in datasets:
+#    if len(dataset) < max_entries:
+#        max_entries = len(dataset)
+
+# Use the first dataset as the benchmark for the number of entries
+# max_entries = len(datasets[0])
+
+# Fuck you, set it manually
+max_entries = 5000
+
+processed_datasets = []
+for dataset, multiplier in datasets:
+    if multiplier is not None:
+        dataset = dataset.select(range(int(max_entries * multiplier)))
+    processed_datasets.append(dataset)
+
+print(f"Datasets truncated to {max_entries} entries")
+
+# Concatenate all datasets
+dataset = concatenate_datasets(processed_datasets)
+
+print("Final dataset size:", len(dataset))
+
+# Shuffle the dataset
+dataset = dataset.shuffle(seed=42)
+
+dataset.push_to_hub("allura-forge/expr-rp-sft-mix")
+
+# Also upload this file as source code
+with open(__file__, "rb") as f:
+    HfApi().upload_file(
+        path_or_fileobj=f,
+        path_in_repo="main.py",
+        repo_id="allura-forge/expr-rp-sft-mix",
+        repo_type="dataset"
+    )