Downloads · 30 days
12
3% of all-time downloads
gpt-imdb-ipo_annealing
Myashka/gpt-imdb-ipo_annealing
gpt-imdb-ipo_annealing is a text generation model from Myashka. Use it when you need the model to write or continue text. It is set up for transformers.
should probably proofread and complete it, then remove this comment. --
All-time downloads
433
Public
Parameters
124M
7 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors498 MB · 99%
At a glance
- Task
- Text Generation
- Library
- transformers
- Model type
- gpt2
- Access
- Public
- Created
- Dec 8, 2023
- Updated
- Dec 9, 2023
- SHA
- 9f993dd5
Try a prompt
Base models
Model card
From the Hugging Face model README
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
gpt-imdb-ipo_annealing
This model is a fine-tuned version of lvwerra/gpt2-imdb on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 125.6974
- Rewards/chosen: -0.0343
- Rewards/rejected: -0.1277
- Rewards/accuracies: 0.875
- Rewards/margins: 0.0934
- Logps/rejected: -267.1282
- Logps/chosen: -236.1897
- Logits/rejected: -31.3501
- Logits/chosen: -31.5916
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 24
- eval_batch_size: 24
- seed: 42
- optimizer: Adam with betas=(0.9,0.99) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 150
- training_steps: 7197
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 16.3187 | 0.21 | 500 | 34.0876 | 0.1161 | -0.1126 | 0.5292 | 0.2287 | -263.8062 | -235.1407 | -33.1877 | -33.4371 |
| 5.5155 | 0.42 | 1000 | 13.0423 | -0.1485 | -0.3812 | 0.5042 | 0.2327 | -264.1273 | -235.4375 | -35.2608 | -35.4541 |
| 10.2532 | 0.63 | 1500 | 18.5157 | -0.4407 | -0.5471 | 0.5458 | 0.1064 | -264.3746 | -235.8205 | -34.2230 | -34.4246 |
| 6.755 | 0.83 | 2000 | 28.1593 | -0.7791 | -0.8052 | 0.5917 | 0.0261 | -264.7961 | -236.3400 | -33.6119 | -33.8069 |
| 9.4126 | 1.04 | 2500 | 9.2406 | -0.8733 | -1.2564 | 0.6229 | 0.3831 | -265.6003 | -236.5962 | -31.9471 | -32.0700 |
| 8.5908 | 1.25 | 3000 | 12.4967 | -0.6700 | -1.0163 | 0.6167 | 0.3462 | -265.4156 | -236.4061 | -31.6914 | -31.8443 |
| 19.5217 | 1.46 | 3500 | 6.8889 | -0.0720 | -0.4689 | 0.6854 | 0.3969 | -264.5895 | -235.4041 | -32.1300 | -32.2692 |
| 6.9195 | 1.67 | 4000 | 4.2435 | -0.5324 | -0.9335 | 0.7021 | 0.4012 | -265.7609 | -236.4489 | -31.8342 | -31.9606 |
| 4.6993 | 1.88 | 4500 | 5.0987 | -0.2002 | -0.6179 | 0.7521 | 0.4177 | -265.3070 | -235.7907 | -31.6301 | -31.7617 |
| 2.7896 | 2.08 | 5000 | 2.7344 | -0.2390 | -0.5589 | 0.7500 | 0.3199 | -265.4754 | -236.0307 | -31.9650 | -32.1009 |
| 3.2262 | 2.29 | 5500 | 3.0584 | -0.1936 | -0.5168 | 0.8083 | 0.3231 | -265.8080 | -236.0606 | -31.6585 | -31.8243 |
| 4.1965 | 2.5 | 6000 | 4.2350 | -0.1555 | -0.4440 | 0.8417 | 0.2884 | -266.2272 | -236.1557 | -31.6484 | -31.8344 |
| 15.1482 | 2.71 | 6500 | 10.8174 | -0.0932 | -0.3244 | 0.8667 | 0.2312 | -266.7491 | -236.1454 | -31.4600 | -31.6800 |
| 145.9251 | 2.92 | 7000 | 125.6974 | -0.0343 | -0.1277 | 0.875 | 0.0934 | -267.1282 | -236.1897 | -31.3501 | -31.5916 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.1
- Datasets 2.15.0
- Tokenizers 0.15.0
- Task
- Text Generation
- Library
- transformers
- Type
- gpt2
- Created
- Dec 8, 2023
- Updated
- Dec 9, 2023