metadata
license: apache-2.0
library_name: peft
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
- 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: 0.6171
- Rewards/chosen: -0.4648
- Rewards/rejected: -0.8388
- Rewards/accuracies: 0.3711
- Rewards/margins: 0.3740
- Logps/rejected: -161.0705
- Logps/chosen: -110.3948
- Logits/rejected: 1.0411
- Logits/chosen: 0.9868
- Use Label: 0.0
- Pred Label: 0.0
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: 8
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- total_eval_batch_size: 64
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6553 | 0.21 | 100 | 0.6557 | -0.1267 | -0.2685 | 0.3633 | 0.1419 | -104.0477 | -76.5787 | -2.0726 | -2.0833 | 0.0 | 0.0 |
| 0.6446 | 0.42 | 200 | 0.6343 | -0.2873 | -0.5376 | 0.3828 | 0.2503 | -130.9503 | -92.6377 | -0.6864 | -0.7124 | 0.0 | 0.0 |
| 0.6273 | 0.63 | 300 | 0.6204 | -0.4623 | -0.7994 | 0.3672 | 0.3371 | -157.1332 | -110.1469 | 0.6726 | 0.6280 | 0.0 | 0.0 |
| 0.6165 | 0.84 | 400 | 0.6182 | -0.4457 | -0.8122 | 0.3672 | 0.3666 | -158.4149 | -108.4784 | 0.9580 | 0.9035 | 0.0 | 0.0 |
Framework versions
- PEFT 0.7.1
- Transformers 4.38.2
- Pytorch 2.1.1+cu121
- Datasets 2.14.6
- Tokenizers 0.15.2