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-2-lr1e-6-e1
    results: []

Llama0-3-8b-v0.1-p-2-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: 0.0114
Rewards/chosen: -180.9830
Rewards/rejected: -175.0705
Rewards/accuracies: 0.4395
Rewards/margins: -5.9126
Logps/rejected: -17593.8047
Logps/chosen: -18186.5996
Logits/rejected: -7.0208
Logits/chosen: -7.0530

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
0.1446	0.2137	100	0.0699	-18.5094	-17.6715	0.4919	-0.8378	-1853.9094	-1939.2302	-1.3162	-1.3101
0.014	0.4275	200	0.0129	-163.4562	-159.6935	0.4556	-3.7627	-16056.1084	-16433.9180	-5.9717	-5.9933
0.0162	0.6412	300	0.0155	-112.5662	-107.6031	0.4516	-4.9631	-10847.0625	-11344.9082	-6.2349	-6.2804
0.0108	0.8549	400	0.0121	-183.4940	-177.6753	0.4435	-5.8188	-17854.2852	-18437.6973	-7.0250	-7.0544

Framework versions

Transformers 4.45.1
Pytorch 2.4.1+cu121
Datasets 3.0.0
Tokenizers 0.20.0