Llama-2-7b-hf-DPO-PartialEval_ET0.1_MT1.2_V.1.0
This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.7063
- Rewards/chosen: -1.8087
- Rewards/rejected: -2.4327
- Rewards/accuracies: 0.7000
- Rewards/margins: 0.6240
- Logps/rejected: -101.7455
- Logps/chosen: -109.7301
- Logits/rejected: -1.0472
- Logits/chosen: -1.0455
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-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 10
- num_epochs: 3
Training results
Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
---|---|---|---|---|---|---|---|---|---|---|---|
0.6461 | 0.3009 | 68 | 0.6839 | 0.0330 | 0.0082 | 0.6000 | 0.0248 | -77.3366 | -91.3130 | -0.2945 | -0.2863 |
0.9723 | 0.6018 | 136 | 0.6779 | 0.0339 | -0.0486 | 0.7000 | 0.0824 | -77.9042 | -91.3046 | -0.3345 | -0.3256 |
0.6461 | 0.9027 | 204 | 0.6352 | -0.0081 | -0.2128 | 0.8000 | 0.2047 | -79.5466 | -91.7240 | -0.3939 | -0.3854 |
0.2832 | 1.2035 | 272 | 0.5825 | -0.8764 | -1.2440 | 0.7000 | 0.3676 | -89.8586 | -100.4076 | -0.6262 | -0.6198 |
0.1923 | 1.5044 | 340 | 0.5559 | -1.1573 | -1.6161 | 0.7000 | 0.4587 | -93.5792 | -103.2166 | -0.6844 | -0.6797 |
0.3898 | 1.8053 | 408 | 0.6173 | -1.3556 | -1.8473 | 0.7000 | 0.4918 | -95.8919 | -105.1990 | -0.8939 | -0.8905 |
0.3404 | 2.1062 | 476 | 0.6381 | -1.3063 | -1.8875 | 0.7000 | 0.5812 | -96.2932 | -104.7061 | -0.9068 | -0.9042 |
0.4954 | 2.4071 | 544 | 0.6915 | -1.7445 | -2.3721 | 0.7000 | 0.6276 | -101.1399 | -109.0883 | -1.0304 | -1.0288 |
0.3914 | 2.7080 | 612 | 0.7063 | -1.8087 | -2.4327 | 0.7000 | 0.6240 | -101.7455 | -109.7301 | -1.0472 | -1.0455 |
Framework versions
- PEFT 0.10.0
- Transformers 4.40.2
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1
- Downloads last month
- 8
Model tree for LBK95/Llama-2-7b-hf-DPO-PartialEval_ET0.1_MT1.2_1-10_V.1.0_Filtered0.2_V1.0
Base model
meta-llama/Llama-2-7b-hf
Adapter
this model