metadata
license: llama2
base_model: meta-llama/Llama-2-7b-hf
tags:
- trl
- dpo
- generated_from_trainer
library_name: peft
model-index:
- name: Llama-2-7b-hf-DPO-LookAhead-5_Q2_TTree1.4_TT0.9_TP0.7_TE0.2_V1
results: []
Llama-2-7b-hf-DPO-LookAhead-5_Q2_TTree1.4_TT0.9_TP0.7_TE0.2_V1
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: 1.2865
- Rewards/chosen: -2.3072
- Rewards/rejected: -1.8542
- Rewards/accuracies: 0.5
- Rewards/margins: -0.4531
- Logps/rejected: -121.6291
- Logps/chosen: -184.9154
- Logits/rejected: 0.2000
- Logits/chosen: 0.1668
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.7102 | 0.3017 | 70 | 0.6900 | 0.0340 | 0.0258 | 0.7000 | 0.0082 | -102.8295 | -161.5030 | 0.6083 | 0.5744 |
| 0.7024 | 0.6034 | 140 | 0.7276 | 0.0806 | 0.1382 | 0.3000 | -0.0575 | -101.7058 | -161.0370 | 0.6015 | 0.5681 |
| 0.6653 | 0.9052 | 210 | 0.7362 | 0.0303 | 0.0739 | 0.4000 | -0.0435 | -102.3490 | -161.5399 | 0.6196 | 0.5858 |
| 0.488 | 1.2069 | 280 | 0.8450 | -0.5919 | -0.4398 | 0.4000 | -0.1521 | -107.4859 | -167.7624 | 0.5482 | 0.5148 |
| 0.5839 | 1.5086 | 350 | 0.8971 | -0.8481 | -0.6497 | 0.4000 | -0.1984 | -109.5843 | -170.3242 | 0.5183 | 0.4846 |
| 0.503 | 1.8103 | 420 | 1.0273 | -1.1487 | -0.8225 | 0.4000 | -0.3262 | -111.3127 | -173.3304 | 0.4207 | 0.3883 |
| 0.2083 | 2.1121 | 490 | 1.1693 | -1.6401 | -1.2436 | 0.4000 | -0.3965 | -115.5236 | -178.2447 | 0.2902 | 0.2576 |
| 0.1395 | 2.4138 | 560 | 1.2310 | -2.1881 | -1.7991 | 0.6000 | -0.3890 | -121.0787 | -183.7240 | 0.2345 | 0.2015 |
| 0.1618 | 2.7155 | 630 | 1.2865 | -2.3072 | -1.8542 | 0.5 | -0.4531 | -121.6291 | -184.9154 | 0.2000 | 0.1668 |
Framework versions
- PEFT 0.12.0
- Transformers 4.44.0
- Pytorch 2.4.0+cu121
- Datasets 3.2.0
- Tokenizers 0.19.1