ValueError: optimizer got an empty parameter list when using group normalization instead of batch normalization in yolov5

[vardanagarwal]

Search before asking

• I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

I am trying to use group normalization as the batch size I can use is small due to memory constraints. To do that I am changing the code in common.py.

class Conv(nn.Module):
    # Standard convolution
    def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True):  # ch_in, ch_out, kernel, stride, padding, groups
        super().__init__()
        self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g, bias=False)
        self.bn = nn.BatchNorm2d(c2)
        self.act = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())

    def forward(self, x):
        return self.act(self.bn(self.conv(x)))

    def forward_fuse(self, x):
        return self.act(self.conv(x))

In this code, the only change done is self.bn = nn.BatchNorm2d(c2) to self.bn = nn.GroupNorm(8, c2)

Now, when trying to run with the command: python3 train.py --data data/coco128.yaml --weights '' --cfg models/yolov5s.yaml --hyp data/hyps/hyp.scratch-low.yaml --epochs 300 --batch 16 --img 640, I get this error.

Traceback (most recent call last):
  File "train.py", line 643, in <module>
    main(opt)
  File "train.py", line 539, in main
    train(opt.hyp, opt, device, callbacks)
  File "train.py", line 170, in train
    optimizer = SGD(g0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True)
  File "/home/dori/env3.8/lib/python3.8/site-packages/torch/optim/sgd.py", line 101, in __init__
    super(SGD, self).__init__(params, defaults)
  File "/home/dori/env3.8/lib/python3.8/site-packages/torch/optim/optimizer.py", line 49, in __init__
    raise ValueError("optimizer got an empty parameter list")
ValueError: optimizer got an empty parameter list

Additional

No response

[github-actions]

👋 Hello @vardanagarwal, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset images, training logs, screenshots, and a public link to online W&B logging if available.

For business inquiries or professional support requests please visit https://ultralytics.com or email support@ultralytics.com.

Requirements

Python>=3.7.0 with all requirements.txt installed including PyTorch>=1.7. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab and Kaggle notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training (train.py), validation (val.py), inference (detect.py) and export (export.py) on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

@vardanagarwal might want to try debugging at a smaller level before running a full model
https://pytorch.org/docs/stable/generated/torch.nn.GroupNorm.html

[glenn-jocher]

@vardanagarwal seems like this is due to LR assignment groups here:

yolov5/train.py

Lines 153 to 161 in 71685cb

	g0, g1, g2 = [], [], [] # optimizer parameter groups
	for v in model.modules():
	if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter): # bias
	g2.append(v.bias)
	if isinstance(v, nn.BatchNorm2d): # weight (no decay)
	g0.append(v.weight)
	elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter): # weight (with decay)
	g1.append(v.weight)

[glenn-jocher]

@vardanagarwal good news 😃! Your original issue may now be fixed ✅ in PR #7376. This PR reorganizes optimizer parameter group inits for robustness to missing BatchNorm2d layers in a model. Your use-case trains correctly now, but note that your normalization layer weights will now see weight decay, which is probably contrary to best practices. You can update train.py to include any custom normalization layers in group 0, which is excluded from weight decay.

To receive this update:

• Git – git pull from within your yolov5/ directory or git clone https://github.com/ultralytics/yolov5 again
• PyTorch Hub – Force-reload model = torch.hub.load('ultralytics/yolov5', 'yolov5s', force_reload=True)
• Notebooks – View updated notebooks
• Docker – sudo docker pull ultralytics/yolov5:latest to update your image

Thank you for spotting this issue and informing us of the problem. Please let us know if this update resolves the issue for you, and feel free to inform us of any other issues you discover or feature requests that come to mind. Happy trainings with YOLOv5 🚀!

[vardanagarwal]

@glenn-jocher thanks for the help, I am able to fix it by adding the group normalization layer.
if isinstance(v, nn.BatchNorm2d) or isinstance(v, nn.GroupNorm):

[glenn-jocher]

@vardanagarwal ah yes, this is a good solution also. Can you please submit a PR to include GroupNorm and any other applicable normalization layer on this line?

Please see our ✅ Contributing Guide to get started.

[vardanagarwal]

Sure, I'll do that.