qwen3_4_20250818_1941

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: 0.3340
• Map@3: 0.9375

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
16.5429	0.0523	20	1.4557	0.7283
9.2865	0.1046	40	0.9736	0.8026
8.3639	0.1569	60	1.0031	0.7932
7.1456	0.2092	80	0.7150	0.8585
6.1949	0.2615	100	0.6272	0.8776
5.3446	0.3138	120	0.6454	0.8768
4.9297	0.3661	140	0.6001	0.8850
4.2539	0.4184	160	0.6017	0.8870
4.9359	0.4707	180	0.5601	0.8877
4.0852	0.5230	200	0.5453	0.8985
4.2137	0.5754	220	0.4796	0.9097
4.1494	0.6277	240	0.4894	0.9105
4.1857	0.6800	260	0.4618	0.9078
3.5215	0.7323	280	0.4672	0.9093
4.2297	0.7846	300	0.4450	0.9139
3.2632	0.8369	320	0.4476	0.9171
4.0446	0.8892	340	0.4467	0.9141
3.4267	0.9415	360	0.4137	0.9207
3.4374	0.9938	380	0.4655	0.9113
3.1897	1.0445	400	0.4886	0.9167
2.413	1.0968	420	0.4331	0.9232
2.7002	1.1491	440	0.4092	0.9242
2.7209	1.2014	460	0.3857	0.9278
2.6897	1.2537	480	0.4045	0.9260
2.3799	1.3060	500	0.3872	0.9310
2.7859	1.3583	520	0.4151	0.9229
2.6904	1.4106	540	0.3789	0.9313
2.4114	1.4629	560	0.3901	0.9302
2.6539	1.5152	580	0.3838	0.9330
2.4441	1.5675	600	0.3571	0.9348
2.086	1.6198	620	0.3667	0.9341
2.0958	1.6721	640	0.3498	0.9375
2.3942	1.7244	660	0.3753	0.9288
2.7639	1.7767	680	0.3384	0.9377
2.2673	1.8290	700	0.3267	0.9380
2.2347	1.8813	720	0.3378	0.9371
2.1848	1.9336	740	0.3271	0.9376
2.1091	1.9859	760	0.3330	0.9369
1.8355	2.0366	780	0.3340	0.9375

Framework versions

• PEFT 0.17.0
• Transformers 4.55.2
• Pytorch 2.6.0+cu124
• Datasets 4.0.0
• Tokenizers 0.21.4