metadata
license: mit
library_name: peft
tags:
- trl
- dpo
- generated_from_trainer
base_model: google/gemma-2b
model-index:
- name: gemma-7b-lora-distilabel-intel-orca-dpo-pairs
results: []
gemma-7b-lora-distilabel-intel-orca-dpo-pairs
This model is a fine-tuned version of google/gemma-2b on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.4641
- Rewards/chosen: -0.2842
- Rewards/rejected: -2.0677
- Rewards/accuracies: 0.8414
- Rewards/margins: 1.7835
- Logps/rejected: -294.6812
- Logps/chosen: -246.1420
- Logits/rejected: -29.7875
- Logits/chosen: -27.6122
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: 1e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 250
- num_epochs: 1
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6333 | 0.19 | 250 | 0.5221 | -0.4649 | -1.1235 | 0.8196 | 0.6586 | -285.2397 | -247.9492 | -29.5102 | -27.3832 |
| 0.4697 | 0.39 | 500 | 0.4819 | -0.5572 | -2.0261 | 0.8394 | 1.4689 | -294.2652 | -248.8721 | -29.5979 | -27.4182 |
| 0.4471 | 0.58 | 750 | 0.4814 | -0.5104 | -2.3183 | 0.8418 | 1.8079 | -297.1878 | -248.4040 | -29.6888 | -27.5182 |
| 0.4477 | 0.78 | 1000 | 0.4744 | -0.3874 | -2.2429 | 0.8418 | 1.8555 | -296.4334 | -247.1736 | -29.7387 | -27.5680 |
| 0.458 | 0.97 | 1250 | 0.4641 | -0.2842 | -2.0677 | 0.8414 | 1.7835 | -294.6812 | -246.1420 | -29.7875 | -27.6122 |
Framework versions
- PEFT 0.8.2
- Transformers 4.38.0
- Pytorch 2.1.0+cu121
- Datasets 2.17.0
- Tokenizers 0.15.2
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 46.47 |
| AI2 Reasoning Challenge (25-Shot) | 49.15 |
| HellaSwag (10-Shot) | 71.78 |
| MMLU (5-Shot) | 41.52 |
| TruthfulQA (0-shot) | 33.10 |
| Winogrande (5-shot) | 65.98 |
| GSM8k (5-shot) | 17.29 |