metadata
library_name: transformers
license: llama3.1
base_model: OpenLeecher/llama_8b_lima_11
tags:
- trl
- dpo
- llama-factory
- generated_from_trainer
model-index:
- name: llama_8b_lima_11_dpo
results: []
llama_8b_lima_11_dpo
This model is a fine-tuned version of OpenLeecher/llama_8b_lima_11 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5851
- Rewards/chosen: -0.4629
- Rewards/rejected: -7.1735
- Rewards/accuracies: 0.9500
- Rewards/margins: 6.7106
- Logps/rejected: -539.3364
- Logps/chosen: -1185.7501
- Logits/rejected: -1.0965
- Logits/chosen: -1.1222
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: 2e-06
- train_batch_size: 1
- eval_batch_size: 1
- seed: 66
- distributed_type: multi-GPU
- num_devices: 2
- gradient_accumulation_steps: 10
- total_train_batch_size: 20
- total_eval_batch_size: 2
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: polynomial
- lr_scheduler_warmup_steps: 20
- 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 |
---|---|---|---|---|---|---|---|---|---|---|---|
0.7617 | 0.2972 | 90 | 0.7245 | -0.7682 | -5.6994 | 0.9000 | 4.9312 | -490.1999 | -1195.9271 | -1.1052 | -1.1403 |
0.7151 | 0.5945 | 180 | 0.6198 | -0.4894 | -6.6033 | 0.9100 | 6.1140 | -520.3312 | -1186.6332 | -1.1077 | -1.1366 |
0.6943 | 0.8917 | 270 | 0.5851 | -0.4629 | -7.1735 | 0.9500 | 6.7106 | -539.3364 | -1185.7501 | -1.0965 | -1.1222 |
Framework versions
- Transformers 4.45.0
- Pytorch 2.4.1+cu124
- Datasets 2.21.0
- Tokenizers 0.20.1