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-0.025-lr5e-7-e1
results: []
Llama0-3-8b-v0.1-p-0.025-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: 0.6532
- Rewards/chosen: -0.5100
- Rewards/rejected: -0.5735
- Rewards/accuracies: 0.5887
- Rewards/margins: 0.0635
- Logps/rejected: -144.1059
- Logps/chosen: -139.2881
- Logits/rejected: 0.1540
- Logits/chosen: 0.1389
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6771 | 0.2137 | 100 | 0.6768 | -0.0374 | -0.0438 | 0.5685 | 0.0063 | -91.1355 | -92.0372 | 0.0615 | 0.0416 |
| 0.6683 | 0.4275 | 200 | 0.6671 | -0.1690 | -0.1914 | 0.6129 | 0.0224 | -105.8998 | -105.1957 | 0.1030 | 0.0845 |
| 0.6575 | 0.6412 | 300 | 0.6582 | -0.3685 | -0.4154 | 0.6089 | 0.0470 | -128.3006 | -125.1377 | 0.1500 | 0.1335 |
| 0.6558 | 0.8549 | 400 | 0.6541 | -0.4822 | -0.5419 | 0.5927 | 0.0597 | -140.9456 | -136.5091 | 0.1504 | 0.1348 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.4.1+cu121
- Datasets 3.0.0
- Tokenizers 0.20.0