metadata
license: apache-2.0
library_name: peft
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
base_model: mistralai/Mistral-7B-v0.1
datasets:
- HuggingFaceH4/ultrafeedback_binarized
model-index:
- name: zephyr-7b
results: []
zephyr-7b
This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-qlora on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:
- Loss: 0.6906
- Rewards/chosen: -0.3413
- Rewards/rejected: -0.5652
- Rewards/accuracies: 0.3631
- Rewards/margins: 0.2239
- Logps/rejected: -131.9189
- Logps/chosen: -103.0295
- Logits/rejected: -0.1381
- Logits/chosen: -0.2453
- Use Label: 15879.8574
- Pred Label: 4192.1431
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6818 | 0.1 | 100 | 0.6814 | -0.0056 | -0.0496 | 0.3393 | 0.0440 | -80.3582 | -69.4632 | -2.0664 | -2.0975 | 1833.4603 | 22.5397 |
| 0.6818 | 0.21 | 200 | 0.6861 | -0.1358 | -0.2381 | 0.3373 | 0.1023 | -99.2068 | -82.4782 | -1.9938 | -2.0215 | 3701.2063 | 258.7936 |
| 0.6848 | 0.31 | 300 | 0.6877 | -0.2068 | -0.3388 | 0.3413 | 0.1320 | -109.2766 | -89.5763 | -1.8828 | -1.9157 | 5437.8730 | 626.1270 |
| 0.6857 | 0.42 | 400 | 0.6885 | -0.1802 | -0.3299 | 0.3532 | 0.1497 | -108.3913 | -86.9237 | -1.4031 | -1.4529 | 7112.4443 | 1055.5555 |
| 0.6894 | 0.52 | 500 | 0.6892 | -0.2862 | -0.4559 | 0.3552 | 0.1697 | -120.9922 | -97.5203 | -0.5997 | -0.6889 | 8741.4287 | 1530.5714 |
| 0.6881 | 0.63 | 600 | 0.6918 | -0.3826 | -0.6059 | 0.3532 | 0.2233 | -135.9845 | -107.1618 | -0.2548 | -0.3579 | 10293.6826 | 2082.3174 |
| 0.6913 | 0.73 | 700 | 0.6899 | -0.3542 | -0.5787 | 0.3671 | 0.2244 | -133.2637 | -104.3247 | -0.2462 | -0.3470 | 11806.4766 | 2673.5239 |
| 0.6893 | 0.84 | 800 | 0.6904 | -0.3443 | -0.5684 | 0.3631 | 0.2241 | -132.2416 | -103.3355 | -0.1293 | -0.2367 | 13331.9043 | 3252.0952 |
| 0.689 | 0.94 | 900 | 0.6907 | -0.3413 | -0.5651 | 0.3631 | 0.2238 | -131.9111 | -103.0301 | -0.1367 | -0.2437 | 14866.4766 | 3821.5239 |
Framework versions
- PEFT 0.7.1
- Transformers 4.38.2
- Pytorch 2.1.1+cu121
- Datasets 2.14.6
- Tokenizers 0.15.2