An error occurred during fine-tuning：TypeError: loss_by_feat() missing 1 required positional argument: 'batch_img_metas'

[liuhy269]

I followed the official documentation to fine-tune the model, but the following error occurred: TypeError: loss_by_feat() missing 1 required positional argument: 'batch_img_metas'. Can anyone tell me how to resolve this?

[JoyRiderJie]

I followed the official documentation to fine-tune the model, but the following error occurred: TypeError: loss_by_feat() missing 1 required positional argument: 'batch_img_metas'. Can anyone tell me how to resolve this?

This error #646 (comment) also occurred when I performed prompt_tuning_coco using the configuration file configs/prompt_tuning_coco/yolo_world_v2_l_vlpan_bn_2e-4_80e_8gpus_mask-refine_prompt_tuning_coco.py.

[JoyRiderJie]

I think the cause of this error is a missing parameter passing. In other words, the loss() function only expects two parameters, but three parameters are being passed here. I'm not sure if this is a bug.

[Winwan]

May I ask if you have solved it?

[JoyRiderJie]

May I ask if you have solved it?
You can refer to this.

[inarticular]

请问您解决这个问题了吗？
您可以参考这个链接。

Hi, thanks for your suggestion.

I modified the code to:

loss_inputs = outs + (txt_masks, batch_data_samples['bboxes_labels'], batch_data_samples['img_metas'])

but I still get the following error:

TypeError: loss() missing 1 required positional argument: 'batch_data_samples'

If it is convenient for you, could you please share your complete yolo_world_v2_l_vlpan_bn_2e-4_80e_8gpus_mask-refine_prompt_tuning_coco.py files for reference? I would really appreciate it, as it would help me compare them with my current version and better understand what else may need to be adjusted.

Thank you again for your time and help.

[AnnicePlayer]

@inarticular hi I also encountered the same problem. I think it might be associated with the setting of the simple detector. Have you solved it?

[AnnicePlayer]

@inarticular actually, I do have a simple fix method, in the line 168 of the yolo world.py, simply change the
losses = self.bbox_head.loss(img_feats, txt_feats, batch_data_samples)

into
losses = self.bbox_head.loss(img_feats, txt_feats,None, batch_data_samples)
to make sure there is no keyword missing.

But it's ridiculous to set txt_masks=None. Are the txt_masks necessary in this simple detector?