qwen3_4_20250809

This model is a fine-tuned version of Qwen/Qwen3-4B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4858
• Map@3: 0.9101

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: 0.0002
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Map@3
21.087	0.0623	20	1.4444	0.7169
9.8108	0.1246	40	1.1465	0.7859
7.6633	0.1868	60	0.9294	0.8218
7.0727	0.2491	80	0.7713	0.8514
6.1375	0.3114	100	0.6833	0.8680
5.8015	0.3737	120	0.6946	0.8675
4.9316	0.4360	140	0.6652	0.8726
5.0053	0.4982	160	0.6030	0.8796
4.5137	0.5605	180	0.5770	0.8903
4.4611	0.6228	200	0.6356	0.8779
4.8127	0.6851	220	0.5487	0.8906
4.5154	0.7474	240	0.5269	0.9004
3.9116	0.8097	260	0.5962	0.8855
4.3977	0.8719	280	0.4977	0.9028
3.6962	0.9342	300	0.4942	0.9084
3.2734	0.9965	320	0.4548	0.9119
3.2019	1.0561	340	0.4396	0.9157
3.1092	1.1183	360	0.5063	0.9056
3.0406	1.1806	380	0.4858	0.9101

Framework versions

• PEFT 0.17.0
• Transformers 4.55.0
• Pytorch 2.6.0+cu124
• Datasets 4.0.0
• Tokenizers 0.21.4