ui-tars-1.5-7b-idm-full-sft-12-frames-videos

This model is a fine-tuned version of Qwen/Qwen2.5-VL-7B-Instruct on the ui_tars_sft_data_12_frames_videos dataset. It achieves the following results on the evaluation set:

• Loss: 0.0279

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: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 16
• total_eval_batch_size: 8
• 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: 0.05
• num_epochs: 6.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.0477	1.0	511	0.0454
0.0176	2.0	1022	0.0271
0.0097	3.0	1533	0.0255
0.0081	4.0	2044	0.0212
0.0046	5.0	2555	0.0241
0.0022	6.0	3066	0.0279

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.0.2
• Tokenizers 0.21.1