metadata
license: llama3.1
library_name: peft
tags:
- trl
- dpo
- generated_from_trainer
base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
model-index:
- name: llama3.1_8b_dpo_bwgenerator_test
results: []
llama3.1_8b_dpo_bwgenerator_test
This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0381
- Rewards/chosen: -9.3770
- Rewards/rejected: -40.9760
- Rewards/accuracies: 0.9961
- Rewards/margins: 31.5990
- Logps/rejected: -519.9075
- Logps/chosen: -178.3189
- Logits/rejected: -1.4901
- Logits/chosen: -1.9907
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: 5e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0864 | 0.0719 | 1000 | 0.1031 | -24.3451 | -55.7071 | 0.9919 | 31.3620 | -667.2187 | -328.0001 | -1.3920 | -1.9101 |
| 0.0721 | 0.1438 | 2000 | 0.0666 | -17.1956 | -43.6489 | 0.9932 | 26.4533 | -546.6367 | -256.5046 | -1.3146 | -1.8819 |
| 0.0513 | 0.2157 | 3000 | 0.0586 | -13.4148 | -39.7394 | 0.9932 | 26.3247 | -507.5419 | -218.6962 | -1.5754 | -2.0549 |
| 0.0391 | 0.2876 | 4000 | 0.0518 | -11.9859 | -42.5627 | 0.9942 | 30.5768 | -535.7746 | -204.4073 | -1.5376 | -2.0293 |
| 0.0431 | 0.3595 | 5000 | 0.0584 | -15.0281 | -51.9022 | 0.9945 | 36.8741 | -629.1698 | -234.8300 | -1.5020 | -2.0037 |
| 0.0386 | 0.4313 | 6000 | 0.0399 | -10.5384 | -39.9545 | 0.9961 | 29.4161 | -509.6927 | -189.9328 | -1.5356 | -2.0315 |
| 0.0417 | 0.5032 | 7000 | 0.0452 | -11.8813 | -46.2602 | 0.9955 | 34.3789 | -572.7493 | -203.3616 | -1.4399 | -1.9551 |
| 0.06 | 0.5751 | 8000 | 0.0387 | -9.4865 | -39.5614 | 0.9958 | 30.0749 | -505.7617 | -179.4136 | -1.5289 | -2.0209 |
| 0.0478 | 0.6470 | 9000 | 0.0376 | -9.9444 | -40.6988 | 0.9961 | 30.7544 | -517.1356 | -183.9923 | -1.5154 | -2.0106 |
| 0.022 | 0.7189 | 10000 | 0.0399 | -9.6813 | -41.9896 | 0.9961 | 32.3084 | -530.0439 | -181.3615 | -1.4896 | -1.9912 |
| 0.0254 | 0.7908 | 11000 | 0.0378 | -9.1448 | -40.6698 | 0.9961 | 31.5250 | -516.8457 | -175.9964 | -1.5031 | -2.0023 |
| 0.0357 | 0.8627 | 12000 | 0.0387 | -9.6321 | -41.6962 | 0.9961 | 32.0641 | -527.1096 | -180.8692 | -1.4851 | -1.9878 |
| 0.0626 | 0.9346 | 13000 | 0.0381 | -9.3770 | -40.9760 | 0.9961 | 31.5990 | -519.9075 | -178.3189 | -1.4901 | -1.9907 |
Framework versions
- PEFT 0.10.0
- Transformers 4.44.0
- Pytorch 2.3.0+cu121
- Datasets 2.14.7
- Tokenizers 0.19.1