--- license: apache-2.0 base_model: HuggingFaceTB/SmolLM-135M-Instruct tags: - trl - orpo - generated_from_trainer model-index: - name: ft-orpo-smollm-135M-instruct-on-hf-ultrafeedback results: [] --- # ft-orpo-smollm-135M-instruct-on-hf-ultrafeedback This model is a fine-tuned version of [HuggingFaceTB/SmolLM-135M-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM-135M-Instruct) on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set: - Loss: 1.1646 - Rewards/chosen: -0.1296 - Rewards/rejected: -0.1298 - Rewards/accuracies: 0.4000 - Rewards/margins: 0.0002 - Logps/rejected: -1.2981 - Logps/chosen: -1.2964 - Logits/rejected: 31.6875 - Logits/chosen: 31.3425 - Nll Loss: 1.0873 - Log Odds Ratio: -0.7727 - Log Odds Chosen: -0.0238 ## 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: 0.0003 - train_batch_size: 4 - eval_batch_size: 4 - seed: 42 - gradient_accumulation_steps: 2 - total_train_batch_size: 8 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - 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 | Nll Loss | Log Odds Ratio | Log Odds Chosen | |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:--------:|:--------------:|:---------------:| | 1.4274 | 0.27 | 100 | 1.2066 | -0.1351 | -0.1347 | 0.4100 | -0.0004 | -1.3467 | -1.3508 | 28.6347 | 28.3442 | 1.1292 | -0.7736 | -0.0347 | | 1.1351 | 0.53 | 200 | 1.1796 | -0.1316 | -0.1316 | 0.4100 | 0.0000 | -1.3162 | -1.3158 | 31.1292 | 30.7764 | 1.1024 | -0.7723 | -0.0251 | | 1.135 | 0.8 | 300 | 1.1646 | -0.1296 | -0.1298 | 0.4000 | 0.0002 | -1.2981 | -1.2964 | 31.6875 | 31.3425 | 1.0873 | -0.7727 | -0.0238 | ### Framework versions - Transformers 4.39.3 - Pytorch 2.1.2 - Datasets 2.18.0 - Tokenizers 0.15.2