cutelemonlili/Qwen2.5-14B-Instruct_lora_MATH_training_response_Qwen2.5_14B_only_right

MATH_training_response_Qwen2.5_14B_only_right

This model is a fine-tuned version of Qwen/Qwen2.5-14B-Instruct on the MATH_training_response_Qwen2.5_14B_only_right dataset. It achieves the following results on the evaluation set:

• Loss: 0.0381

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
0.0514	0.1404	200	0.0501
0.0505	0.2807	400	0.0442
0.0339	0.4211	600	0.0449
0.0418	0.5614	800	0.0442
0.0518	0.7018	1000	0.0431
0.0376	0.8421	1200	0.0409
0.0311	0.9825	1400	0.0377
0.0251	1.1228	1600	0.0393
0.0386	1.2632	1800	0.0397
0.034	1.4035	2000	0.0392
0.023	1.5439	2200	0.0391
0.0204	1.6842	2400	0.0382
0.0079	1.8246	2600	0.0381
0.0192	1.9649	2800	0.0378

Framework versions

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