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-lr1e-6-e1
results: []
Llama0-3-8b-v0.1-dpo-lr1e-6-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: 0.6081
- Rewards/chosen: -1.5097
- Rewards/rejected: -1.7233
- Rewards/accuracies: 0.6129
- Rewards/margins: 0.2136
- Logps/rejected: -259.0845
- Logps/chosen: -239.2623
- Logits/rejected: 0.4158
- Logits/chosen: 0.4094
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: 1e-06
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.652 | 0.2137 | 100 | 0.6498 | -0.3163 | -0.3568 | 0.6169 | 0.0405 | -122.4376 | -119.9265 | 0.1437 | 0.1272 |
| 0.6273 | 0.4275 | 200 | 0.6269 | -0.8960 | -1.0115 | 0.6129 | 0.1155 | -187.9061 | -177.8965 | 0.3135 | 0.3044 |
| 0.6152 | 0.6412 | 300 | 0.6155 | -1.2887 | -1.4649 | 0.6129 | 0.1762 | -233.2506 | -217.1634 | 0.4062 | 0.4005 |
| 0.608 | 0.8549 | 400 | 0.6094 | -1.4301 | -1.6328 | 0.6089 | 0.2027 | -250.0380 | -231.3053 | 0.4062 | 0.4004 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.4.1+cu121
- Datasets 3.0.0
- Tokenizers 0.20.0