--- base_model: Minbyul/llama2-7b-wo-kqa_silver_wogold-sft tags: - trl - dpo - generated_from_trainer model-index: - name: llama2-7b-dpo-full-sft-wo-kqa_silver_wogold results: [] --- # llama2-7b-dpo-full-sft-wo-kqa_silver_wogold This model is a fine-tuned version of [Minbyul/llama2-7b-wo-kqa_silver_wogold-sft](https://huggingface.co/Minbyul/llama2-7b-wo-kqa_silver_wogold-sft) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.4456 - Rewards/chosen: -0.0307 - Rewards/rejected: -1.1443 - Rewards/accuracies: 0.9418 - Rewards/margins: 1.1136 - Logps/rejected: -714.0911 - Logps/chosen: -108.8464 - Logits/rejected: -0.4045 - Logits/chosen: -0.9202 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:| | 0.2705 | 0.93 | 100 | 0.4456 | -0.0307 | -1.1443 | 0.9418 | 1.1136 | -714.0911 | -108.8464 | -0.4045 | -0.9202 | ### Framework versions - Transformers 4.39.0.dev0 - Pytorch 2.1.2 - Datasets 2.14.6 - Tokenizers 0.15.2