model_usp1_dpo9
This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.4153
- Rewards/chosen: -3.7856
- Rewards/rejected: -9.9056
- Rewards/accuracies: 0.7700
- Rewards/margins: 6.1200
- Logps/rejected: -125.7940
- Logps/chosen: -114.6437
- Logits/rejected: 0.0370
- Logits/chosen: 0.1165
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: 0.0005
- train_batch_size: 4
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.1142 | 2.67 | 100 | 1.4766 | 1.7195 | -0.5069 | 0.6900 | 2.2263 | -115.3510 | -108.5270 | -0.1562 | -0.1401 |
| 0.0431 | 5.33 | 200 | 2.4525 | -23.9043 | -28.0492 | 0.6600 | 4.1448 | -145.9536 | -136.9979 | -0.5122 | -0.4496 |
| 0.0035 | 8.0 | 300 | 2.2089 | -1.4351 | -6.8781 | 0.75 | 5.4430 | -122.4302 | -112.0322 | 0.1830 | 0.2540 |
| 0.0 | 10.67 | 400 | 2.4382 | -3.7206 | -9.8637 | 0.7700 | 6.1431 | -125.7475 | -114.5715 | 0.0396 | 0.1191 |
| 0.0 | 13.33 | 500 | 2.4513 | -3.7668 | -9.8629 | 0.7700 | 6.0960 | -125.7465 | -114.6229 | 0.0387 | 0.1181 |
| 0.0 | 16.0 | 600 | 2.4281 | -3.7498 | -9.8885 | 0.7700 | 6.1387 | -125.7750 | -114.6039 | 0.0379 | 0.1174 |
| 0.0 | 18.67 | 700 | 2.4461 | -3.7862 | -9.9007 | 0.7700 | 6.1145 | -125.7885 | -114.6444 | 0.0376 | 0.1165 |
| 0.0 | 21.33 | 800 | 2.4172 | -3.7467 | -9.9054 | 0.7700 | 6.1587 | -125.7938 | -114.6005 | 0.0372 | 0.1166 |
| 0.0 | 24.0 | 900 | 2.4260 | -3.7798 | -9.9215 | 0.7700 | 6.1417 | -125.8117 | -114.6373 | 0.0377 | 0.1171 |
| 0.0 | 26.67 | 1000 | 2.4153 | -3.7856 | -9.9056 | 0.7700 | 6.1200 | -125.7940 | -114.6437 | 0.0370 | 0.1165 |
Framework versions
- PEFT 0.10.0
- Transformers 4.39.3
- Pytorch 2.2.2+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 2
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
Model tree for guoyu-zhang/model_usp1_dpo9
Base model
meta-llama/Llama-2-7b-chat-hf