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-p-0.05-lr5e-7-e1
results: []
Llama0-3-8b-v0.1-p-0.05-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: 0.6417
- Rewards/chosen: -0.5291
- Rewards/rejected: -0.5936
- Rewards/accuracies: 0.5968
- Rewards/margins: 0.0645
- Logps/rejected: -146.1117
- Logps/chosen: -141.1991
- Logits/rejected: 0.1599
- Logits/chosen: 0.1449
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6655 | 0.2137 | 100 | 0.6653 | -0.0372 | -0.0437 | 0.5847 | 0.0065 | -91.1306 | -92.0131 | 0.0596 | 0.0396 |
| 0.6569 | 0.4275 | 200 | 0.6557 | -0.1710 | -0.1939 | 0.6089 | 0.0230 | -106.1501 | -105.3901 | 0.1033 | 0.0850 |
| 0.646 | 0.6412 | 300 | 0.6467 | -0.3795 | -0.4276 | 0.6008 | 0.0480 | -129.5149 | -126.2473 | 0.1520 | 0.1356 |
| 0.6441 | 0.8549 | 400 | 0.6425 | -0.4991 | -0.5612 | 0.5887 | 0.0622 | -142.8797 | -138.2009 | 0.1546 | 0.1390 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.4.1+cu121
- Datasets 3.0.0
- Tokenizers 0.20.0