This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 0.5772
Rewards/chosen: -1.0430
Rewards/rejected: -1.6270
Rewards/accuracies: 0.7063
Rewards/margins: 0.5841
Logps/rejected: -558.6804
Logps/chosen: -509.0577
Logits/rejected: -2.5324
Logits/chosen: -2.3779

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-07
train_batch_size: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 4
total_train_batch_size: 128
total_eval_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6351	0.2093	100	0.6359	-0.6754	-0.9179	0.6746	0.2426	-487.7697	-472.2982	-2.4565	-2.2922
0.6101	0.4186	200	0.5990	-0.7996	-1.1966	0.7143	0.3970	-515.6393	-484.7244	-2.4477	-2.2933
0.5738	0.6279	300	0.5819	-1.0722	-1.6607	0.7143	0.5885	-562.0454	-511.9821	-2.5003	-2.3506
0.5808	0.8373	400	0.5776	-1.0426	-1.6196	0.7063	0.5769	-557.9310	-509.0269	-2.6060	-2.4454

Framework versions

Transformers 4.44.2
Pytorch 2.4.0+cu121
Datasets 2.19.1
Tokenizers 0.19.1

tanliboy/lambda-llama-3-8b-dpo-test

lambda-llama-3-8b-dpo-test

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tanliboy/
lambda-llama-3-8b-dpo-test