UWNSL/Qwen2.5-32B-Instruct_Long_CoT_lora

MATH_training_QwQ_32B_Preview

This model is a fine-tuned version of Qwen/Qwen2.5-32B-Instruct on the MATH_training_Qwen_QwQ_32B_Preview dataset. It achieves the following results on the evaluation set:

• Loss: 0.1193

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 4
• total_eval_batch_size: 4
• optimizer: Use 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: 2.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.2322	0.1564	200	0.2143
0.1419	0.3127	400	0.1674
0.1137	0.4691	600	0.1510
0.1457	0.6255	800	0.1406
0.0953	0.7819	1000	0.1333
0.1201	0.9382	1200	0.1268
0.0899	1.0946	1400	0.1289
0.0548	1.2510	1600	0.1269
0.0323	1.4073	1800	0.1240
0.0414	1.5637	2000	0.1207
0.0375	1.7201	2200	0.1202
0.0512	1.8765	2400	0.1201

Framework versions

• PEFT 0.12.0
• Transformers 4.46.1
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3