Makrrr/qwen2.5-7B-reasonmed-finetune-extreme

reason_med_final

This model is a fine-tuned version of ../models/Qwen2.5-7B-Instruct on the reason_med_full dataset. It achieves the following results on the evaluation set:

• Loss: 0.4633

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• total_eval_batch_size: 8
• 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: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.514	0.3844	1000	0.4880
0.4958	0.5766	1500	0.4770
0.4856	0.7688	2000	0.4683
0.4809	0.9610	2500	0.4610
0.4048	1.1534	3000	0.4647
0.4073	1.3456	3500	0.4587
0.3971	1.5378	4000	0.4527
0.3985	1.7300	4500	0.4479
0.402	1.9222	5000	0.4420
0.3008	2.1142	5500	0.4667
0.2989	2.3064	6000	0.4659
0.2955	2.4986	6500	0.4652
0.2926	2.6908	7000	0.4641
0.2982	2.8830	7500	0.4634

Framework versions

• Transformers 4.52.4
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1