qwen25_7b_rm_acc_5e5_3ep

This model is a fine-tuned version of Jennny/qwen25_7b_sft_math on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1814
• Accuracy: 0.72

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-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 256
• total_eval_batch_size: 32
• 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: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.1609	0.4848	10	0.9237	0.6467
0.2258	0.9697	20	0.8203	0.65
0.1396	1.4848	30	0.7778	0.66
0.1603	1.9697	40	0.8515	0.6867
0.0226	2.4848	50	1.2602	0.7067
0.011	2.9697	60	1.1814	0.72

Framework versions

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