metadata
license: apache-2.0
library_name: peft
tags:
- trl
- dpo
- generated_from_trainer
base_model: mistralai/Mistral-7B-v0.1
model-index:
- name: zephyr-7b
results: []
zephyr-7b
This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.3625
- Rewards/chosen: -150.6127
- Rewards/rejected: -146.3050
- Rewards/accuracies: 0.2421
- Rewards/margins: -4.3077
- Logps/rejected: -14705.8975
- Logps/chosen: -15130.1680
- Logits/rejected: 13.5362
- Logits/chosen: 13.4716
- Use Label: 11165.9844
- Pred Label: 7522.0161
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-06
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- 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 | Use Label | Pred Label |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6637 | 0.1 | 100 | 0.6642 | -0.0947 | -0.1635 | 0.3254 | 0.0687 | -91.7446 | -78.3734 | -2.0927 | -2.1253 | 1838.9207 | 17.0794 |
| 0.3902 | 0.21 | 200 | 0.3930 | -14.4219 | -14.0352 | 0.2560 | -0.3866 | -1478.9202 | -1511.0870 | 2.8471 | 2.7727 | 3444.6985 | 515.3016 |
| 0.3845 | 0.31 | 300 | 0.3786 | -23.0869 | -24.5685 | 0.2520 | 1.4817 | -2532.2498 | -2377.5872 | 5.4283 | 5.3070 | 4579.4922 | 1484.5079 |
| 0.3477 | 0.42 | 400 | 0.3622 | -111.3259 | -109.5294 | 0.25 | -1.7965 | -11028.3408 | -11201.4893 | 11.6816 | 11.5716 | 5682.4922 | 2485.5081 |
| 0.3468 | 0.52 | 500 | 0.3613 | -144.7782 | -140.7408 | 0.2421 | -4.0373 | -14149.4824 | -14546.7158 | 13.8885 | 13.8347 | 6784.2383 | 3487.7620 |
| 0.33 | 0.63 | 600 | 0.3605 | -143.0167 | -138.8336 | 0.2401 | -4.1831 | -13958.7627 | -14370.5693 | 12.5943 | 12.5399 | 7857.4287 | 4518.5713 |
| 0.3665 | 0.73 | 700 | 0.3614 | -150.1877 | -145.8865 | 0.2421 | -4.3011 | -14664.0518 | -15087.6680 | 13.4024 | 13.3367 | 8936.4287 | 5543.5713 |
| 0.3731 | 0.84 | 800 | 0.3623 | -150.4385 | -146.1303 | 0.2401 | -4.3082 | -14688.4258 | -15112.7539 | 13.5339 | 13.4696 | 10050.3330 | 6533.6665 |
| 0.3696 | 0.94 | 900 | 0.3625 | -150.6127 | -146.3050 | 0.2421 | -4.3077 | -14705.8975 | -15130.1680 | 13.5362 | 13.4716 | 11165.9844 | 7522.0161 |
Framework versions
- PEFT 0.7.1
- Transformers 4.38.2
- Pytorch 2.1.1+cu121
- Datasets 2.14.6
- Tokenizers 0.15.2