423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-14 10:56:56 +08:00
Code Issues Packages Projects Releases Wiki Activity

refactor dataset_attr, add eos in pt, fix #757

Browse Source 
Former-commit-id: a9d1fb72f7
This commit is contained in:
hiyouga
2023-09-01 19:00:45 +08:00
parent 173022473d
commit a4fd976048
20 changed files with 160 additions and 198 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
data/README.md
View File

@@ -6,13 +6,13 @@ If you are using a custom dataset, please provide your dataset definition in the
"script_url": "the name of the directory containing a dataset loading script. (if specified, ignore below 2 arguments)",
"file_name": "the name of the dataset file in the this directory. (required if above are not specified)",
"file_sha1": "the SHA-1 hash value of the dataset file. (optional)",
"ranking": "whether the examples contains ranked responses or not. (default: false)",
"columns": {
"prompt": "the name of the column in the datasets containing the prompts. (default: instruction)",
"query": "the name of the column in the datasets containing the queries. (default: input)",
"response": "the name of the column in the datasets containing the responses. (default: output)",
"history": "the name of the column in the datasets containing the history of chat. (default: None)"
},
"stage": "The stage at which the data is being used: pt, sft, and rm, which correspond to pre-training, supervised fine-tuning(PPO), and reward model (DPO) training, respectively.(default: None)"
}
}
```
@@ -27,7 +27,6 @@ For datasets used in reward modeling or DPO training, the `response` column shou
"output": [
"Chosen answer",
"Rejected answer"
],
"stage": "rm"
]
}
```

9
data/README_zh.md
View File

@@ -6,19 +6,19 @@
"script_url": "包含数据加载脚本的本地文件夹名称（若指定，则忽略下列两个参数）",
"file_name": "该目录下数据集文件的名称（若上述参数未指定，则此项必需）",
"file_sha1": "数据集文件的SHA-1哈希值（可选）",
"ranking": "数据集是否包含排序后的回答（默认：false）",
"columns": {
"prompt": "数据集代表提示词的表头名称（默认：instruction）",
"query": "数据集代表请求的表头名称（默认：input）",
"response": "数据集代表回答的表头名称（默认：output）",
"history": "数据集代表历史对话的表头名称（默认：None）"
},
"stage": "数据所应用的训练阶段，可选值有 pt, sft, rm 三个，对应预训练，指令监督微调(PPO)，奖励模型(DPO)训练, 默认为None，表示不限制"
}
}
```
其中 `prompt` 和 `response` 列应当是非空的字符串。`query` 列的内容将会和 `prompt` 列拼接作为模型输入。`history` 列应当是一个列表，其中每个元素是一个字符串二元组，分别代表用户请求和模型答复。
对于奖励模型或 DPO 训练的数据集，`response` 列应当是一个字符串列表，排在前面的代表更优的答案，例如：
对于训练奖励模型或 DPO 训练的数据集，`response` 列应当是一个字符串列表，排在前面的代表更优的答案，例如：
```json
{
@@ -27,7 +27,6 @@
"output": [
"Chosen answer",
"Rejected answer"
],
"stage": "rm"
]
}
```

124
data/dataset_info.json
View File

@@ -1,28 +1,23 @@
{
"alpaca_en": {
"file_name": "alpaca_data_en_52k.json",
"file_sha1": "607f94a7f581341e59685aef32f531095232cf23",
"stage": "sft"
"file_sha1": "607f94a7f581341e59685aef32f531095232cf23"
},
"alpaca_zh": {
"file_name": "alpaca_data_zh_51k.json",
"file_sha1": "e655af3db557a4197f7b0cf92e1986b08fae6311",
"stage": "sft"
"file_sha1": "e655af3db557a4197f7b0cf92e1986b08fae6311"
},
"alpaca_gpt4_en": {
"file_name": "alpaca_gpt4_data_en.json",
"file_sha1": "647f4ad447bd993e4b6b6223d1be15208bab694a",
"stage": "sft"
"file_sha1": "647f4ad447bd993e4b6b6223d1be15208bab694a"
},
"alpaca_gpt4_zh": {
"file_name": "alpaca_gpt4_data_zh.json",
"file_sha1": "3eaa3bda364ccdd59925d7448a698256c31ef845",
"stage": "sft"
"file_sha1": "3eaa3bda364ccdd59925d7448a698256c31ef845"
},
"self_cognition": {
"file_name": "self_cognition.json",
"file_sha1": "6287a730ada924fc5d9eadc6d8f865e01b7a6f67",
"stage": "sft"
"file_sha1": "6287a730ada924fc5d9eadc6d8f865e01b7a6f67"
},
"oaast_sft": {
"file_name": "oaast_sft.json",
@@ -32,8 +27,7 @@
"query": "input",
"response": "output",
"history": "history"
},
"stage": "sft"
}
},
"oaast_sft_zh": {
"file_name": "oaast_sft_zh.json",
@@ -43,8 +37,7 @@
"query": "input",
"response": "output",
"history": "history"
},
"stage": "sft"
}
},
"sharegpt_zh": {
"file_name": "sharegpt_zh_27k.json",
@@ -54,8 +47,7 @@
"query": "input",
"response": "output",
"history": "history"
},
"stage": "sft"
}
},
"lima": {
"file_name": "lima.json",
@@ -65,8 +57,7 @@
"query": "input",
"response": "output",
"history": "history"
},
"stage": "sft"
}
},
"example": {
"script_url": "example_dataset",
@@ -75,32 +66,25 @@
"query": "input",
"response": "output",
"history": "history"
},
"stage": "sft"
}
},
"guanaco": {
"hf_hub_url": "JosephusCheung/GuanacoDataset",
"stage": "sft"
"hf_hub_url": "JosephusCheung/GuanacoDataset"
},
"belle_0.5m": {
"hf_hub_url": "BelleGroup/train_0.5M_CN",
"stage": "sft"
"hf_hub_url": "BelleGroup/train_0.5M_CN"
},
"belle_1m": {
"hf_hub_url": "BelleGroup/train_1M_CN",
"stage": "sft"
"hf_hub_url": "BelleGroup/train_1M_CN"
},
"belle_2m": {
"hf_hub_url": "BelleGroup/train_2M_CN",
"stage": "sft"
"hf_hub_url": "BelleGroup/train_2M_CN"
},
"belle_dialog": {
"hf_hub_url": "BelleGroup/generated_chat_0.4M",
"stage": "sft"
"hf_hub_url": "BelleGroup/generated_chat_0.4M"
},
"belle_math": {
"hf_hub_url": "BelleGroup/school_math_0.25M",
"stage": "sft"
"hf_hub_url": "BelleGroup/school_math_0.25M"
},
"belle_multiturn": {
"script_url": "belle_multiturn",
@@ -109,8 +93,7 @@
"query": "",
"response": "output",
"history": "history"
},
"stage": "sft"
}
},
"firefly": {
"hf_hub_url": "YeungNLP/firefly-train-1.1M",
@@ -119,16 +102,13 @@
"query": "",
"response": "target",
"history": ""
},
"stage": "sft"
}
},
"codealpaca": {
"hf_hub_url": "sahil2801/CodeAlpaca-20k",
"stage": "sft"
"hf_hub_url": "sahil2801/CodeAlpaca-20k"
},
"alpaca_cot": {
"hf_hub_url": "QingyiSi/Alpaca-CoT",
"stage": "sft"
"hf_hub_url": "QingyiSi/Alpaca-CoT"
},
"webqa": {
"hf_hub_url": "suolyer/webqa",
@@ -137,8 +117,7 @@
"query": "",
"response": "output",
"history": ""
},
"stage": "sft"
}
},
"ultra_chat": {
"script_url": "ultra_chat",
@@ -147,32 +126,29 @@
"query": "",
"response": "output",
"history": "history"
},
"stage": "sft"
}
},
"novel_tokens512_50k": {
"hf_hub_url": "zxbsmk/webnovel_cn",
"stage": "sft"
"hf_hub_url": "zxbsmk/webnovel_cn"
},
"ad_gen": {
"adgen": {
"hf_hub_url": "HasturOfficial/adgen",
"columns": {
"prompt": "content",
"query": "",
"response": "summary",
"history": ""
},
"stage": "sft"
}
},
"comparison_gpt4_en": {
"file_name": "comparison_gpt4_data_en.json",
"file_sha1": "96fa18313544e22444fe20eead7754b17da452ae",
"stage": "rm"
"ranking": true
},
"comparison_gpt4_zh": {
"file_name": "comparison_gpt4_data_zh.json",
"file_sha1": "515b18ed497199131ddcc1af950345c11dc5c7fd",
"stage": "rm"
"ranking": true
},
"hh_rlhf_en": {
"script_url": "hh_rlhf_en",
@@ -182,7 +158,7 @@
"response": "output",
"history": "history"
},
"stage": "rm"
"ranking": true
},
"oaast_rm": {
"file_name": "oaast_rm.json",
@@ -193,7 +169,7 @@
"response": "output",
"history": "history"
},
"stage": "rm"
"ranking": true
},
"oaast_rm_zh": {
"file_name": "oaast_rm_zh.json",
@@ -204,7 +180,7 @@
"response": "output",
"history": "history"
},
"stage": "rm"
"ranking": true
},
"wiki_demo": {
"file_name": "wiki_demo.txt",
@@ -214,8 +190,7 @@
"query": "",
"response": "",
"history": ""
},
"stage": "pt"
}
},
"refinedweb": {
"hf_hub_url": "tiiuae/falcon-refinedweb",
@@ -224,18 +199,7 @@
"query": "",
"response": "",
"history": ""
},
"stage": "pt"
},
"starcoder": {
"hf_hub_url": "bigcode/starcoderdata",
"columns": {
"prompt": "content",
"query": "",
"response": "",
"history": ""
},
"stage": "pt"
}
},
"wikipedia_en": {
"hf_hub_url": "olm/olm-wikipedia-20221220",
@@ -244,8 +208,7 @@
"query": "",
"response": "",
"history": ""
},
"stage": "pt"
}
},
"wikipedia_zh": {
"hf_hub_url": "pleisto/wikipedia-cn-20230720-filtered",
@@ -254,7 +217,24 @@
"query": "",
"response": "",
"history": ""
},
"stage": "pt"
}
},
"the_stack": {
"hf_hub_url": "bigcode/the-stack",
"columns": {
"prompt": "content",
"query": "",
"response": "",
"history": ""
}
},
"starcoder": {
"hf_hub_url": "bigcode/starcoderdata",
"columns": {
"prompt": "content",
"query": "",
"response": "",
"history": ""
}
}
}