This model is a fine-tuned version of W-61/llama-3-8b-base-sft-ultrachat-8xh200 on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 1.2244
Rewards/chosen: -0.0088
Rewards/rejected: -0.0106
Rewards/accuracies: 0.6028
Rewards/margins: 0.0018
Logps/rejected: -1.0607
Logps/chosen: -0.8758
Logits/rejected: -0.4873
Logits/chosen: -0.5022
Nll Loss: 1.2174
Log Odds Ratio: -0.6565
Log Odds Chosen: 0.2857

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: 8
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 4
total_train_batch_size: 128
total_eval_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
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	Nll Loss	Log Odds Ratio	Log Odds Chosen
5.2395	0.4188	200	1.2644	-0.0088	-0.0107	0.6008	0.0018	-1.0672	-0.8826	-0.4935	-0.5068	1.2561	-0.6568	0.2836
4.9046	0.8377	400	1.2244	-0.0088	-0.0106	0.6028	0.0018	-1.0607	-0.8758	-0.4873	-0.5022	1.2174	-0.6565	0.2857

Framework versions

Transformers 4.51.0
Pytorch 2.3.1+cu121
Datasets 2.21.0
Tokenizers 0.21.4

jackf857/llama-3-8b-base-orpo-ultrafeedback-4xh200-rerun

llama-3-8b-base-orpo-ultrafeedback-4xh200-rerun

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

jackf857/
llama-3-8b-base-orpo-ultrafeedback-4xh200-rerun