ModuleNotFoundError: No module named 'pathlib._local'; 'pathlib' is not a package

[Obrigad0]

Search before asking

• I have searched the YOLOv5 issues and found no similar bug report.

YOLOv5 Component

Validation

Bug

Hello YOLOv5 team,

I am experiencing a persistent ModuleNotFoundError: No module named 'models' when trying to validate my model using val.py in the original YOLOv5 repository folder. Here are all the details of my setup and what I have tried so far:

Environment:

What I am doing:

I trained a model on the CCPD2019 dataset using YOLOv5 in its original folder.

I am always working inside the original YOLOv5 repository folder (where val.py, train.py, and the models/ directory are located).

I have not created any custom folders or modules named models.

I have no pip packages named models installed (pip show models returns nothing).

My working directory is always the root of the YOLOv5 repo.

I am running the following command:

bash
python val.py --weights runs/train/exp15/weights/best.pt --data ./data/ccpd.yaml --img 640 --save-json
Error message:

text
ModuleNotFoundError: No module named 'models'
This happens during the torch.load() step when loading the weights.

What I have tried:

Verified that there are no conflicting folders or packages named models in my project or installed via pip.

Checked that my working directory is correct and contains the original YOLOv5 structure.

Confirmed that my Python and PyTorch versions are the same as during training.

Tried adding sys.path.insert(0, './yolov5') at the top of my scripts, but since I am already in the repo root, this should not be necessary.

Uninstalled any unnecessary packages like pathlib that could cause conflicts.

Attempted to extract the state_dict from the checkpoint using a workaround, but the error occurs before I can access the checkpoint.

Additional info:

The model was trained and saved in the same environment and folder where I am now trying to validate it.

The error persists even after restarting the environment.

Request:
Could you please help me troubleshoot this issue? Is there something I am missing regarding the environment, repo structure, or model saving/loading process?
Any suggestions or insights would be greatly appreciated!

Thank you very much for your support.

Let me know if you want to add or change any detail!

Environment

Python version: 3.11.7

PyTorch version: 2.3.0+cu121

CUDA: NVIDIA GeForce RTX 4090, 24GB

YOLOv5 version: v7.0-421-g79c4c31d (latest from GitHub)

Minimal Reproducible Example

literally the code provided of val.py from github, there were no changes (like the train file used for training)

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[UltralyticsAssistant]

👋 Hello @Obrigad0, thank you for your detailed report and for using YOLOv5 🚀! This is an automated response to help you get started. An Ultralytics engineer will review your issue and assist you further soon.

If this is a 🐛 Bug Report, please provide a minimum reproducible example (MRE) if you haven't already. This helps us quickly diagnose and resolve the issue.

For general troubleshooting, please ensure you have followed our Tutorials, including Custom Data Training and validation steps. Also, review our Tips for Best Training Results to ensure your environment matches our recommendations.

Requirements

Python>=3.8.0 with all requirements.txt installed, including PyTorch>=1.8. To set up your environment:

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

Environments

YOLOv5 is verified to work in the following environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• 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, validation, inference, export and benchmarks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

🐞 If you have additional details or a more concise MRE, please update your issue. Thank you for your patience and for helping us improve YOLOv5!

[pderrenger]

The error in your issue title suggests a pathlib._local import issue, which is different from the models module error you describe in the details. This typically indicates an environment contamination where a custom pathlib package may have been installed that conflicts with Python's built-in pathlib module.

Please try these steps to resolve the issue:

1. Check for conflicting pathlib installations:

   pip list | grep pathlib
   pip uninstall pathlib pathlib2

2. Verify your environment is clean:

   python -c "import pathlib; print(pathlib.__file__)"

   This should show a path within your Python installation, not a site-packages path.

3. Update to the latest YOLOv5 version:

   git pull origin master
   pip install -U -r requirements.txt

4. Test with a fresh Python environment:

   python -m venv fresh_env
   source fresh_env/bin/activate  # On Windows: fresh_env\Scripts\activate
   pip install -r requirements.txt

The pathlib._local error usually resolves once any conflicting pathlib packages are removed from your environment.

[SisroCodeke]

add this before your code

import os
import sys
import pathlib

sys.modules["pathlib._local"] = pathlib 
if os.name == 'nt':
   pathlib.PosixPath = pathlib.WindowsPath
else:
   pathlib.WindowsPath = pathlib.PosixPath

first saw it here :
#13578

this answer worked for me :
#13578 (comment)

tested on:
intel nuc11
i5-1135G7
Pop_OS (Ubuntu 22.04)
python3.10/python3.12

[pderrenger]

Thanks for sharing the workaround @SisroCodeke! While that patch can work temporarily, I'd still recommend properly fixing the environment by removing conflicting pathlib installations as this addresses the root cause rather than masking the symptom.