rtmdet-ins training error :TypeError: loss_by_feat() takes from 5 to 6 positional arguments but 7 were given

[W-hary]

Prerequisite

• [X] I have searched the existing and past issues but cannot get the expected help.
• [X] I have read the FAQ documentation but cannot get the expected help.
• [X] The bug has not been fixed in the latest version.

🐞 Describe the bug

Install and successfully run the demo test according to ‘15_minutes_instance_segmentation.md‘. An error occurred when training the balloon data set using ‘rtmdet-ins_s_syncbn_fast_8xb32-300e_coco.py’

python tools/train.py configs/rtmdet/rtmdet-ins_s_syncbn_fast_8xb32-300e_coco.py

Traceback (most recent call last): File "tools/train.py", line 123, in main() File "tools/train.py", line 119, in main runner.train() File "/root/miniconda3/lib/python3.8/site-packages/mmengine/runner/runner.py", line 1745, in train model = self.train_loop.run() # type: ignore File "/root/miniconda3/lib/python3.8/site-packages/mmengine/runner/loops.py", line 96, in run self.run_epoch() File "/root/miniconda3/lib/python3.8/site-packages/mmengine/runner/loops.py", line 112, in run_epoch self.run_iter(idx, data_batch) File "/root/miniconda3/lib/python3.8/site-packages/mmengine/runner/loops.py", line 128, in run_iter outputs = self.runner.model.train_step( File "/root/miniconda3/lib/python3.8/site-packages/mmengine/model/base_model/base_model.py", line 114, in train_step losses = self._run_forward(data, mode='loss') # type: ignore File "/root/miniconda3/lib/python3.8/site-packages/mmengine/model/base_model/base_model.py", line 340, in _run_forward results = self(*data, mode=mode) File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl return forward_call(input, *kwargs) File "/root/miniconda3/lib/python3.8/site-packages/mmdet/models/detectors/base.py", line 92, in forward return self.loss(inputs, data_samples) File "/root/miniconda3/lib/python3.8/site-packages/mmdet/models/detectors/single_stage.py", line 78, in loss losses = self.bbox_head.loss(x, batch_data_samples) File "/root/autodl-tmp/mmyolo/mmyolo/models/dense_heads/yolov5_head.py", line 470, in loss losses = self.loss_by_feat(loss_inputs) TypeError: loss_by_feat() takes from 5 to 6 positional arguments but 7 were given

Environment

PyTorch 1.11.0 Python 3.8(ubuntu20.04) Cuda 11.3

Additional information

No response

[collinmccarthy]

This is indirectly because loss_by_feat() isn't implemented for the instance segmentation head. That is, instance segmentation is only supported currently for inference. I don't think it is that difficult to implement this function (you can look at the YOLOv5 detection vs. instance seg versions, and then the RTMDet detection version) but I haven't tried to get it working yet.

If I get it working I'll submit a PR, but I'm not sure when that will be.

Related to #649, #853

[confusedgreenhand]

Hi, I would like to ask if this problem has been solved

[rbli-john]

Same question. Look forward to seeing the fast training for Rtmdet-ins

[jslok]

No update on this?

Am I correct in assuming rtmdet-ins on mmdetection is the same as on mmyolo anyways but just slower training?

One difference I found is mmdeploy only offers deployment to Core ML with segmentation through mmseg.

[collinmccarthy]

Does mmdetection support instance segmentation training with rtmdet? Or just inference like mmyolo? -CollinOn Feb 6, 2024, at 6:19 PM, jlok @.***> wrote: No update on this? Am I correct in assuming rtmdet-ins on mmdetection is the same as on mmyolo anyways but just slower training?

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: @.***>

[jslok]

Mmdet supports training and inference on rtmdet-ins

[HanTnx]

Mmdet supports training and inference on rtmdet-ins

can you provide more information about configuration of mmdet, mmcv, and related mmyolo version ? Thank you

[lijoe123]

The same error with the yolov8_ins, could you please tell me how to fix the problem #636