metadata
library_name: transformers
license: llama3
base_model: meta-llama/Meta-Llama-3-8B-Instruct
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: Llama0-3-8b-v0.1-dpo-lr5e-7-e1
results: []
Llama0-3-8b-v0.1-dpo-lr5e-7-e1
This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.6647
- Rewards/chosen: -0.4944
- Rewards/rejected: -0.5567
- Rewards/accuracies: 0.5968
- Rewards/margins: 0.0624
- Logps/rejected: -142.4301
- Logps/chosen: -137.7279
- Logits/rejected: 0.1487
- Logits/chosen: 0.1334
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-07
- train_batch_size: 2
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 8
- total_train_batch_size: 128
- total_eval_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 1.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6888 | 0.2137 | 100 | 0.6885 | -0.0371 | -0.0441 | 0.5806 | 0.0070 | -91.1662 | -92.0013 | 0.0596 | 0.0400 |
| 0.6799 | 0.4275 | 200 | 0.6786 | -0.1682 | -0.1907 | 0.6089 | 0.0225 | -105.8212 | -105.1108 | 0.1016 | 0.0832 |
| 0.669 | 0.6412 | 300 | 0.6697 | -0.3621 | -0.4081 | 0.6008 | 0.0459 | -127.5619 | -124.5048 | 0.1494 | 0.1334 |
| 0.6673 | 0.8549 | 400 | 0.6657 | -0.4687 | -0.5277 | 0.5887 | 0.0590 | -139.5236 | -135.1644 | 0.1472 | 0.1318 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.4.1+cu121
- Datasets 3.0.0
- Tokenizers 0.20.0