metadata
license: apache-2.0
library_name: peft
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
- trl
- dpo
- alignment-handbook
- 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: 0.6152
- Rewards/chosen: -0.5055
- Rewards/rejected: -0.8740
- Rewards/accuracies: 0.3789
- Rewards/margins: 0.3685
- Logps/rejected: -164.5882
- Logps/chosen: -114.4584
- Logits/rejected: 1.5984
- Logits/chosen: 1.5070
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6551 | 0.21 | 100 | 0.6526 | -0.2364 | -0.3728 | 0.3359 | 0.1364 | -114.4721 | -87.5525 | -1.7460 | -1.7620 | 0.0 | 0.0 |
| 0.6376 | 0.42 | 200 | 0.6289 | -0.3405 | -0.6072 | 0.3672 | 0.2667 | -137.9142 | -97.9614 | 0.0432 | -0.0238 | 0.0 | 0.0 |
| 0.6196 | 0.63 | 300 | 0.6189 | -0.3871 | -0.7293 | 0.375 | 0.3422 | -150.1250 | -102.6218 | 1.1831 | 1.0945 | 0.0 | 0.0 |
| 0.6139 | 0.84 | 400 | 0.6157 | -0.4865 | -0.8500 | 0.3711 | 0.3636 | -162.1976 | -112.5605 | 1.5453 | 1.4533 | 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