TianyiQ commited on
Commit
e2e3c1c
1 Parent(s): fbf290c

Upload ./README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +3 -2
README.md CHANGED
@@ -76,9 +76,9 @@ ProgressGym-HistLlama3-8B-C018-instruct is one of the **36 historical language m
76
  | 1.988 | 3.7158 | 3400 | 2.2770 |
77
  | 2.0355 | 3.9344 | 3600 | 2.2772 |
78
 
79
- Note that the training data volume for the continued pretraining stage is capped at 300MB. When the corresponding century's corpus exceeds this volume, the training data is randomly sampled to fit the volume.
80
 
81
- **ProgressGym-HistLlama3-8B-C018-instruct is an instruction-tuned language model.** It is tuned on [ProgressGym-TimelessQA](https://huggingface.co/datasets/PKU-Alignment/ProgressGym-TimelessQA), using the following hyperparameters:
82
  - learning_rate: 1.5e-05
83
  - train_batch_size: 8
84
  - eval_batch_size: 16
@@ -114,6 +114,7 @@ Note that the training data volume for the continued pretraining stage is capped
114
  - **[Leaderboard & Interactive Playground]** [PKU-Alignment/ProgressGym-LeaderBoard](https://huggingface.co/spaces/PKU-Alignment/ProgressGym-LeaderBoard)
115
  - **[Huggingface Data & Model Collection]** [PKU-Alignment/ProgressGym](https://huggingface.co/collections/PKU-Alignment/progressgym-666735fcf3e4efa276226eaa)
116
  - **[Github Codebase]** [PKU-Alignment/ProgressGym](https://github.com/PKU-Alignment/ProgressGym)
 
117
  - **[PyPI Package]** *(coming soon - [stay tuned](https://forms.gle/1TWFLL4ZCLeYTD5N6)!)*
118
 
119
  ## Citation
 
76
  | 1.988 | 3.7158 | 3400 | 2.2770 |
77
  | 2.0355 | 3.9344 | 3600 | 2.2772 |
78
 
79
+ Note that the training data volume for the continued pretraining stage is capped at 3GB. When the corresponding century's corpus exceeds this volume, the training data is randomly sampled to fit the volume.
80
 
81
+ **ProgressGym-HistLlama3-8B-C018-instruct is an instruction-tuned language model.** It is tuned on [ProgressGym-TimelessQA](https://huggingface.co/datasets/PKU-Alignment/ProgressGym-TimelessQA), using the following hyperparameters. Note, however, that the snapshot at training step 10 is used for the final model, to minimize erosion of the value tendencies learned during continued pretraining; we qualitatively observe that this snapshot still possesses strong instruction-following capabilities.
82
  - learning_rate: 1.5e-05
83
  - train_batch_size: 8
84
  - eval_batch_size: 16
 
114
  - **[Leaderboard & Interactive Playground]** [PKU-Alignment/ProgressGym-LeaderBoard](https://huggingface.co/spaces/PKU-Alignment/ProgressGym-LeaderBoard)
115
  - **[Huggingface Data & Model Collection]** [PKU-Alignment/ProgressGym](https://huggingface.co/collections/PKU-Alignment/progressgym-666735fcf3e4efa276226eaa)
116
  - **[Github Codebase]** [PKU-Alignment/ProgressGym](https://github.com/PKU-Alignment/ProgressGym)
117
+ - **[Documentation]** [ProgressGym Documentation](https://pku-alignment.github.io/ProgressGym/)
118
  - **[PyPI Package]** *(coming soon - [stay tuned](https://forms.gle/1TWFLL4ZCLeYTD5N6)!)*
119
 
120
  ## Citation