how to save and load my custom model about ultralytics HOT 6 OPEN

👋 Hello @iyzyi, thank you for your interest in Ultralytics YOLOv8 🚀! We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Join the vibrant Ultralytics Discord 🎧 community for real-time conversations and collaborations. This platform offers a perfect space to inquire, showcase your work, and connect with fellow Ultralytics users.

Install

Pip install the ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8.

pip install ultralytics

Environments

YOLOv8 may be run in any of the following up-to-date verified 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 Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLOv8 Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

from ultralytics.

model.save("/path") with doc

    Loads parameters from the specified weights file into the model.

    This method supports loading weights from a file or directly from a weights object. It matches parameters by
    name and shape and transfers them to the model.

    Args:
        weights (str | Path): Path to the weights file or a weights object. Defaults to 'yolov8n.pt'.

    Returns:
        self (ultralytics.engine.model.Model): The instance of the class with loaded weights.

    Raises:
        AssertionError: If the model is not a PyTorch model.

from ultralytics.

model.save("/path") with doc

    Loads parameters from the specified weights file into the model.

    This method supports loading weights from a file or directly from a weights object. It matches parameters by
    name and shape and transfers them to the model.

    Args:
        weights (str | Path): Path to the weights file or a weights object. Defaults to 'yolov8n.pt'.

    Returns:
        self (ultralytics.engine.model.Model): The instance of the class with loaded weights.

    Raises:
        AssertionError: If the model is not a PyTorch model.

Thanks.
But my problem is the same as #10297, and I quote:

I then export this as a PyTorch image using model.save('path/to/somefilename.pt'). So these weights are saved locally then.

I then try to load the model with YOLO('path/to/somefilename.pt') and then I get a YOLO model, but it has all the default YOLO classes and not the custom one I used for training any more.

Any idea how to export the model such that when I reload it, the classes are preserved?

from ultralytics.

@iyzyi this should not happen, probably a bug, are you using the latest version of ultralytics?

from ultralytics.

@Kayzwer I'm using 8.2.18.

However it happen even if i upgrade to 8.2.19.

my code:

from ultralytics import YOLO

yolo_model = 'model/yolov8n.pt'
custom_yaml = 'dataset/dataset.yaml'
custom_model = 'model/blue_focus.pt'

model = YOLO(yolo_model)
model.train(data=custom_yaml, epochs=3)
model.save(custom_model)
print('**********************saved**********************')

model = YOLO(custom_model)
model.val()

custom yaml:

path: my_path
train: images/train
val: images/val   
test: images/test

names:
  0: blue_focus

This code will output:

...... (some output) ...... 
Results saved to runs\detect\train14

**********************saved**********************

Dataset 'coco.yaml' images not found ⚠️, missing path 'C:\Users\xxxxx\AppData\Roaming\Ultralytics\datasets\coco\val2017.txt'
Downloading https://github.com/ultralytics/yolov5/releases/download/v1.0/coco2017labels-segments.zip to 'C:\Users\xxxxx\AppData\Roaming\Ultralytics\datasets\coco2017labels-segments.zip'...
100%|██████████| 169M/169M [01:46<00:00, 1.66MB/s]
Unzipping C:\Users\xxxxx\AppData\Roaming\Ultralytics\datasets\coco2017labels-segments.zip to C:\Users\xxxxx\AppData\Roaming\Ultralytics\datasets\coco...: 100%|██████████| 122232/122232 [01:10<00:00, 1727.47file/s]
Downloading http://images.cocodataset.org/zips/train2017.zip to 'C:\Users\xxxxx\AppData\Roaming\Ultralytics\datasets\coco\images\train2017.zip'...
Downloading http://images.cocodataset.org/zips/val2017.zip to 'C:\Users\xxxxx\AppData\Roaming\Ultralytics\datasets\coco\images\val2017.zip'...
Downloading http://images.cocodataset.org/zips/test2017.zip to 'C:\Users\xxxxx\AppData\Roaming\Ultralytics\datasets\coco\images\test2017.zip'...

I never use coco.yaml. It seemed that I get a YOLO model, but it has all the default YOLO classes and not the custom one I used for training any more. Just like #10297

Looking forward to your reply. Thanks!

from ultralytics.

I am sure my dataset is right because I can use bset.pt to do my custom predict.

from ultralytics.