Tensor sizes error when testing a modified YOLOv5 nano model (.yaml)

[lessleeb]

Search before asking

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

Question

Hi, I am trying to modify YOLOv5 nano model. In this case I eliminated the head P5, as shown below in the .yaml file. (it's actually commented).
Additionally, since I am using only 2 anchors now, I commented the corresponding one (P5), and modified the detection part as well. Although, in the training part I do not have any problem, the detection part shows an error regarding the tensor sizes. I have been trying to find what could be the problem, but I've not been able to find it.
This only happens when I remove a layer from the model. If I replace a layer (ex: a Conv for a Focus layer) there is not problem in the detection part.
Could you tell me what could be the problem? Do I need to modify anything else in the code?
Here is the modified yaml file.

YOLOv5n.YAML

Parameters

nc: 80 # number of classes
depth_multiple: 0.33 # model depth multiple
width_multiple: 0.25 # layer channel multiple

anchors:

• [10,13, 16,30, 33,23] # P3/8
• [30,61, 62,45, 59,119] # P4/16
  #- [116,90, 156,198, 373,326] # P5/32

YOLOv5 v6.0 backbone

backbone:

[[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2
[-1, 1, Conv, [128, 3, 2]], # 1-P2/4
[-1, 3, C3, [128]],
[-1, 1, Conv, [256, 3, 2]], # 3-P3/8
[-1, 6, C3, [256]],
[-1, 1, Conv, [512, 3, 2]], # 5-P4/16
[-1, 9, C3, [512]],
[-1, 1, Conv, [1024, 3, 2]], # 7-P5/32
[-1, 3, C3, [1024]],
[-1, 1, SPPF, [1024, 5]], # 9
]

YOLOv5 v6.0 head

head:
[[-1, 1, Conv, [512, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 6], 1, Concat, [1]], # cat backbone P4
[-1, 3, C3, [512, False]], # 13

[-1, 1, Conv, [256, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 4], 1, Concat, [1]], # cat backbone P3
[-1, 3, C3, [256, False]], # 17 (P3/8-small)

[-1, 1, Conv, [256, 3, 2]],
[[-1, 14], 1, Concat, [1]], # cat head P4
[-1, 3, C3, [512, False]], # 20 (P4/16-medium)

[[17, 20], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]

And this is the error:
Exception has occurred: RuntimeError
torch.cat(): Sizes of tensors must match except in dimension 1. Got 26 and 25 in dimension 2 (The offending index is 1)
File "C:\Users\Desktop\yolov5-20211118\models\common.py", line 274, in forward
return torch.cat(x, self.d)
File "C:\Users\Desktop\yolov5-20211118\models\yolo.py", line 150, in _forward_once
x = m(x) # run
File "C:\Users\Desktop\yolov5-20211118\models\yolo.py", line 127, in forward
return self._forward_once(x, profile, visualize) # single-scale inference, train
File "C:\Users\Desktop\yolov5-20211118\models\common.py", line 356, in forward
y = self.model(im) if self.jit else self.model(im, augment=augment, visualize=visualize)
File "C:\Users\Desktop\yolov5-20211118\detect.py", line 115, in run
pred = model(im, augment=augment, visualize=visualize)
File "C:\Users\Desktop\yolov5-20211118\detect.py", line 240, in main
run(**vars(opt))
File "C:\Users\Desktop\yolov5-20211118\detect.py", line 245, in
main(opt)

Additional

No response

[github-actions]

👋 Hello @lessleeb, 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 Glenn Jocher at glenn.jocher@ultralytics.com.

Requirements

Python>=3.6.0 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]

@lessleeb this allows you to create a P3, P4 output YOLOv5s model. Large object performance will be reduced however.

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 80  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.50  # layer channel multiple
anchors: 3

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
   [-1, 3, C3, [128]],
   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
   [-1, 6, C3, [256]],
   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
   [-1, 9, C3, [512]],
   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
   [-1, 3, C3, [1024]],
   [-1, 1, SPPF, [1024, 5]],  # 9
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, Conv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3, [512, False]],  # 13

   [-1, 1, Conv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)

   [-1, 1, Conv, [256, 3, 2]],
   [[-1, 14], 1, Concat, [1]],  # cat head P4
   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)

   [[17, 20], 1, Detect, [nc, anchors]],  # Detect(P3, P4)
  ]

[lessleeb]

Hi @glenn-jocher . Thank you for your response.
I think you meant anchors : 2, right? Since I am using only P3, P4 outputs.
I have modified the .yaml file as you indicated here, and I am still having the same error. Let me explain it.
Train:
Since I am using a customized dataset.
I trained using the pretrained weights weights, default=ROOT / 'yolov5n.pt'
and, cfg, default='models/yolov5n.yaml
Detect:
I used the weights obtained after training.
weights, default=ROOT / 'runs/train/exp/weights/last.pt'

I think when training the model is saved along with the weights, so I don't understand what is the problem when running the detection.

[glenn-jocher]

@lessleeb I created a PR with a -34 model and am able to reproduce your error above now. Seems related to minimum stride constraints dropping from 32 to 16 on this custom model. Might need to enforce a floor of 32, I'll take a look.

[glenn-jocher]

@lessleeb good news 😃! Your original issue may now be fixed ✅ in PR #6230. This PR fixes your original issue and also creates an official P3, P4 only output config yolov5-p34.yaml in models/hub. A quick training and inference verification passes without error:

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 🚀!

[lessleeb]

Thank you! I have updated and tested YOLO in my environment.
There is not error anymore.