LoRA_Qwen2.5_VL_7B_50epoch

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

• Loss: 0.1603

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: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• 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: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
25.8075	1.0	223	3.5207
16.5701	2.0	446	2.3349
9.2497	3.0	669	0.8742
2.2874	4.0	892	0.3473
2.0775	5.0	1115	0.2468
1.6581	6.0	1338	0.2013
1.2788	7.0	1561	0.1920
1.5721	8.0	1784	0.1822
0.9412	9.0	2007	0.1732
1.1306	10.0	2230	0.1694
1.3541	11.0	2453	0.1679
1.3841	12.0	2676	0.1726
0.9869	13.0	2899	0.1602
0.9746	14.0	3122	0.1692
0.9068	15.0	3345	0.1621
0.872	16.0	3568	0.1624
1.1422	17.0	3791	0.1601
1.3074	18.0	4014	0.1598
1.3052	19.0	4237	0.1600
1.1649	20.0	4460	0.1603

Framework versions

• PEFT 0.15.2
• Transformers 4.53.0.dev0
• Pytorch 2.7.1+cu118
• Tokenizers 0.21.1