UWNSL/Qwen2.5-14B-Instruct_Long_CoT_lora

MATH_training_response_QwQ_32B_Preview

This model is a fine-tuned version of Qwen/Qwen2.5-14B-Instruct on the MATH_training_Qwen_QwQ_32B_Preview dataset. It achieves the following results on the evaluation set:

• Loss: 0.2534

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
0.3648	0.1500	200	0.3127
0.2407	0.3001	400	0.2855
0.2113	0.4501	600	0.2747
0.2523	0.6002	800	0.2683
0.2713	0.7502	1000	0.2642
0.2373	0.9002	1200	0.2599
0.1968	1.0503	1400	0.2605
0.2904	1.2003	1600	0.2587
0.1625	1.3503	1800	0.2572
0.2277	1.5004	2000	0.2559
0.2696	1.6504	2200	0.2538
0.2377	1.8005	2400	0.2540
0.1775	1.9505	2600	0.2534

Framework versions

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