selfbiorag-7b-dpo-full-wo-medication_qa-ep3
This model is a fine-tuned version of dmis-lab/selfbiorag_7b on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:
- Loss: 0.6629
- Rewards/chosen: 0.0685
- Rewards/rejected: 0.0097
- Rewards/accuracies: 0.6584
- Rewards/margins: 0.0588
- Logps/rejected: -147.2814
- Logps/chosen: -147.0686
- Logits/rejected: -1.7393
- Logits/chosen: -1.7911
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 |
---|---|---|---|---|---|---|---|---|---|---|---|
0.6575 | 0.25 | 100 | 0.6771 | 0.0687 | 0.0368 | 0.6298 | 0.0319 | -144.5707 | -147.0498 | -1.7114 | -1.7555 |
0.6216 | 0.5 | 200 | 0.6684 | 0.0831 | 0.0335 | 0.6546 | 0.0496 | -144.9030 | -145.6110 | -1.6971 | -1.7417 |
0.6027 | 0.76 | 300 | 0.6636 | 0.0736 | 0.0155 | 0.6536 | 0.0581 | -146.6977 | -146.5513 | -1.7452 | -1.7907 |
Framework versions
- Transformers 4.39.0.dev0
- Pytorch 2.1.2
- Datasets 2.14.6
- Tokenizers 0.15.2
- Downloads last month
- 4