metadata
base_model: Minbyul/selfbiorag-7b-wo-kqa_silver_wogold-sft
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
- trl
- dpo
- generated_from_trainer
datasets:
- HuggingFaceH4/ultrafeedback_binarized
model-index:
- name: selfbiorag-7b-dpo-full-sft-wo-kqa_silver_wogold
results: []
selfbiorag-7b-dpo-full-sft-wo-kqa_silver_wogold
This model is a fine-tuned version of Minbyul/selfbiorag-7b-wo-kqa_silver_wogold-sft on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:
- Loss: 0.3027
- Rewards/chosen: -1.2987
- Rewards/rejected: -5.2535
- Rewards/accuracies: 0.8697
- Rewards/margins: 3.9549
- Logps/rejected: -1323.0533
- Logps/chosen: -663.1464
- Logits/rejected: -0.3508
- Logits/chosen: -0.2498
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: 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.2357 | 0.35 | 100 | 0.3364 | -0.8114 | -3.4524 | 0.8579 | 2.6410 | -1142.9419 | -614.4251 | -0.0367 | 0.0065 |
| 0.154 | 0.71 | 200 | 0.3009 | -0.9868 | -4.1905 | 0.8632 | 3.2037 | -1216.7521 | -631.9589 | -0.3156 | -0.2396 |
Framework versions
- Transformers 4.39.0.dev0
- Pytorch 2.1.2
- Datasets 2.14.6
- Tokenizers 0.15.2