prm_qwen25_math_version3_subsample_hf

This model is a fine-tuned version of Qwen/Qwen2.5-Math-7B-Instruct on the prm_conversations_prm_version3_math+webinstructsub-mcq+webinstructsub-oe+apps+gsm_mix_ref_subsample_hf dataset. It achieves the following results on the evaluation set:

• Loss: 0.1639

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-06
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
0.2317	0.1127	500	0.2231
0.2003	0.2253	1000	0.2006
0.166	0.3380	1500	0.1876
0.1755	0.4506	2000	0.1788
0.1612	0.5633	2500	0.1725
0.1607	0.6759	3000	0.1680
0.1655	0.7886	3500	0.1653
0.1486	0.9012	4000	0.1641

Framework versions

• Transformers 4.45.0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3