biomistral-7b-dpo-full-wo-kqa_silver_wogold-ep3
This model is a fine-tuned version of BioMistral/BioMistral-7B on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:
- Loss: 0.5095
- Rewards/chosen: -0.5198
- Rewards/rejected: -1.3097
- Rewards/accuracies: 0.7480
- Rewards/margins: 0.7899
- Logps/rejected: -355.7885
- Logps/chosen: -252.1618
- Logits/rejected: 0.4827
- Logits/chosen: -1.3398
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-07
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 2
- 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 |
---|---|---|---|---|---|---|---|---|---|---|---|
0.34 | 0.4 | 100 | 0.5716 | -0.2893 | -0.6625 | 0.7202 | 0.3731 | -291.0631 | -229.1131 | 0.0163 | -1.5214 |
0.2189 | 0.79 | 200 | 0.5083 | -0.4815 | -1.2309 | 0.75 | 0.7494 | -347.9087 | -248.3297 | 0.4307 | -1.3596 |
Framework versions
- Transformers 4.39.0.dev0
- Pytorch 2.1.2
- Datasets 2.14.6
- Tokenizers 0.15.2
- Downloads last month
- 3