llama3_8b_helpsteer_correct_rm

This model is a fine-tuned version of Jennny/llama3_8b_sft_helpsteer on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.5813
• Accuracy: 0.5847

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-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• total_eval_batch_size: 16
• optimizer: Use OptimizerNames.PAGED_ADAMW 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: 0.03
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.1608	0.1882	10	1.1481	0.5847
0.128	0.3765	20	0.8165	0.6045
0.1509	0.5647	30	0.7681	0.6271
0.1323	0.7529	40	0.8492	0.5650
0.0856	0.9412	50	0.9250	0.6073
0.0013	1.1129	60	1.2136	0.6073
0.0088	1.3012	70	1.7243	0.5593
0.034	1.4894	80	1.7962	0.5650
0.0083	1.6776	90	1.6045	0.5763
0.0001	1.8659	100	1.5813	0.5847

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.1