library_name, license, base_model, tags, model-index

library_name	license	base_model	tags	model-index
transformers	other	Qwen/Qwen3-4B-Instruct-2507	llama-factory full generated_from_trainer	name results sft_cot

sft_cot

This model is a fine-tuned version of Qwen/Qwen3-4B-Instruct-2507 on the scope_sft_cot dataset. It achieves the following results on the evaluation set:

• Loss: 0.3770

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: 1e-05
• train_batch_size: 1
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_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: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.5282	0.0763	500	0.5320
0.5294	0.1526	1000	0.5253
0.5116	0.2289	1500	0.5039
0.5121	0.3052	2000	0.4850
0.4778	0.3815	2500	0.4666
0.4823	0.4578	3000	0.4533
0.4292	0.5341	3500	0.4343
0.4333	0.6104	4000	0.4167
0.4153	0.6867	4500	0.4017
0.3973	0.7630	5000	0.3908
0.4168	0.8393	5500	0.3823
0.3918	0.9156	6000	0.3777
0.3942	0.9919	6500	0.3770

Framework versions

• Transformers 4.57.1
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.22.1