Predictions on train and val in callbacks

[pzae]

Hello, thank you very much for this excellent library!

I would like to evaluate the model after each training epoch: I want to plot a custom metric versus the number of epochs for both the training and validation sets, and also display visualizations of predictions for a few validation examples to see how the model is learning. I managed to implement this using callbacks, but the code turned out to be rather complex — it includes saving and loading the model as well as obtaining predictions.

My question is: are there any methods in the trainer or validator that already contain the model’s predictions on the training and validation sets, to help simplify my task?

[UltralyticsAssistant]

👋 Hello @pzae, thank you for your interest in Ultralytics 🚀! This is an automated response — an Ultralytics engineer will also assist you here shortly.

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 custom training ❓ Question, please provide as much information as possible, including:

• brief details of your dataset (with a few example images if possible)
• your training logs and environment details
• a minimal snippet of your current callback implementation
  And please verify you are following our Tips for Best Training Results. This will help us understand your workflow and suggest the simplest approach for tracking metrics and visualizations across epochs.

Join the Ultralytics community where it suits you best. For real-time chat, head to Discord 🎧. Prefer in-depth discussions? Check out Discourse. Or dive into threads on our Subreddit to share knowledge with the community.

Upgrade

Upgrade to the latest ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8 to verify your issue is not already resolved in the latest version:

pip install -U ultralytics

Environments

YOLO 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 YOLO Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

Great detail, thanks. The class-name mismatch is from re-instantiating with YOLO(trainer.args.model) and then .load(...), which can preserve the base (COCO) names; instead, use the in-memory weights that already carry your dataset names, or load the epoch checkpoint directly:

import os, glob, torch
from ultralytics import YOLO

def on_fit_epoch_end(trainer):
    # Prefer EMA; already has correct `names` from training
    m = trainer.ema.ema or trainer.model
    img = sorted(glob.glob(os.path.join(trainer.data['val'], '*')))[0]

    m.eval()
    with torch.inference_mode():
        y = YOLO(m)  # no disk I/O, keeps custom classes
        r = y.predict(source=img, imgsz=trainer.args.imgsz, save=False, verbose=False)
        r[0].show()

    # Alternative (uses disk, but also correct classes each epoch):
    # y = YOLO(trainer.last); y.predict(source=img, imgsz=trainer.args.imgsz, save=False, verbose=False)

During training we set model.names from your data, so sticking to trainer.ema.ema/trainer.model preserves your custom classes; see the set_model_attributes behavior in the Trainer reference for details. For how eval/inference is run without gradients, the Model.val() docs outline the same eval-mode, no-grad flow. If errors persist after upgrading, please restart your Python/Jupyter kernel and share yolo checks output plus the full traceback as text.

• Reference: Trainer.set_model_attributes (names are set from data)
• Reference: Model.val() method (eval, no grads)

[pzae]

The error still persists :(

Here is the yolo checks result

(pl_env) E:\python_learning>yolo checks
Ultralytics 8.3.203 Python-3.10.6 torch-2.6.0+cu118 CUDA:0 (NVIDIA GeForce RTX 3060, 12288MiB)
Setup complete (12 CPUs, 31.1 GB RAM, 353.4/465.1 GB disk)

OS Windows-10-10.0.26100-SP0
Environment Windows
Python 3.10.6
Install pip
Path E:\python_learning\pl_env\Lib\site-packages\ultralytics
RAM 31.11 GB
Disk 353.4/465.1 GB
CPU AMD Ryzen 5 7600X 6-Core Processor
CPU count 12
GPU NVIDIA GeForce RTX 3060, 12288MiB
GPU count 1
CUDA 11.8

numpy 2.1.1>=1.23.0
matplotlib 3.10.1>=3.3.0
opencv-python 4.11.0.86>=4.6.0
pillow 11.1.0>=7.1.2
pyyaml 6.0.2>=5.3.1
requests 2.32.3>=2.23.0
scipy 1.15.2>=1.4.1
torch 2.6.0+cu118>=1.8.0
torch 2.6.0+cu118!=2.4.0,>=1.8.0; sys_platform == "win32"
torchvision 0.21.0+cu118>=0.9.0
psutil 7.0.0
polars 1.33.1
ultralytics-thop 2.0.14>=2.0.0

And the...

...error text

FileNotFoundError Traceback (most recent call last) Cell In[4], line 19 17 model = YOLO("yolo11n.pt") 18 model.add_callback("on_fit_epoch_end", on_fit_epoch_end) ---> 19 model.train(data=DATA_YAML, epochs=3)

File E:\python_learning\pl_env\lib\site-packages\ultralytics\engine\model.py:800, in Model.train(self, trainer, **kwargs)
797 self.trainer.model = self.trainer.get_model(weights=self.model if self.ckpt else None, cfg=self.model.yaml)
798 self.model = self.trainer.model
--> 800 self.trainer.train()
801 # Update model and cfg after training
802 if RANK in {-1, 0}:

File E:\python_learning\pl_env\lib\site-packages\ultralytics\engine\trainer.py:235, in BaseTrainer.train(self)
232 ddp_cleanup(self, str(file))
234 else:
--> 235 self._do_train()

File E:\python_learning\pl_env\lib\site-packages\ultralytics\engine\trainer.py:498, in BaseTrainer._do_train(self)
496 self.scheduler.last_epoch = self.epoch # do not move
497 self.stop |= epoch >= self.epochs # stop if exceeded epochs
--> 498 self.run_callbacks("on_fit_epoch_end")
499 self._clear_memory(0.5) # clear if memory utilization > 50%
501 # Early Stopping

File E:\python_learning\pl_env\lib\site-packages\ultralytics\engine\trainer.py:208, in BaseTrainer.run_callbacks(self, event)
206 """Run all existing callbacks associated with a particular event."""
207 for callback in self.callbacks.get(event, []):
--> 208 callback(self)

Cell In[4], line 13, in on_fit_epoch_end(trainer)
11 m.eval()
12 with torch.inference_mode():
---> 13 y = YOLO(m) # no disk I/O, keeps custom classes
14 r = y.predict(source=img, imgsz=trainer.args.imgsz, save=False, verbose=False)
15 r[0].show()

File E:\python_learning\pl_env\lib\site-packages\ultralytics\models\yolo\model.py:83, in YOLO.init(self, model, task, verbose)
80 self.dict = new_instance.dict
81 else:
82 # Continue with default YOLO initialization
---> 83 super().init(model=model, task=task, verbose=verbose)
84 if hasattr(self.model, "model") and "RTDETR" in self.model.model[-1]._get_name(): # if RTDETR head
85 from ultralytics import RTDETR

File E:\python_learning\pl_env\lib\site-packages\ultralytics\engine\model.py:153, in Model.init(self, model, task, verbose)
151 self._new(model, task=task, verbose=verbose)
152 else:
--> 153 self._load(model, task=task)
155 # Delete super().training for accessing self.model.training
156 del self.training

File E:\python_learning\pl_env\lib\site-packages\ultralytics\engine\model.py:302, in Model._load(self, weights, task)
300 self.ckpt_path = self.model.pt_path
301 else:
--> 302 weights = checks.check_file(weights) # runs in all cases, not redundant with above call
303 self.model, self.ckpt = weights, None
304 self.task = task or guess_model_task(weights)

File E:\python_learning\pl_env\lib\site-packages\ultralytics\utils\checks.py:578, in check_file(file, suffix, download, download_dir, hard)
576 files = glob.glob(str(ROOT / "**" / file), recursive=True) or glob.glob(str(ROOT.parent / file)) # find file
577 if not files and hard:
--> 578 raise FileNotFoundError(f"'{file}' does not exist")
579 elif len(files) > 1 and hard:
580 raise FileNotFoundError(f"Multiple files match '{file}', specify exact path: {files}")

FileNotFoundError: 'DetectionModel(
(model): Sequential(
(0): Conv(
(conv): Conv2d(3, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(16, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(16, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(2): C3k2(
(cv1): Conv(
(conv): Conv2d(32, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(48, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): Bottleneck(
(cv1): Conv(
(conv): Conv2d(16, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(8, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(8, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(16, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
(3): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(4): C3k2(
(cv1): Conv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(96, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): Bottleneck(
(cv1): Conv(
(conv): Conv2d(32, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(16, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
(5): Conv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(6): C3k2(
(cv1): Conv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(192, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): C3k(
(cv1): Conv(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv3): Conv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(cv1): Conv(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Bottleneck(
(cv1): Conv(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
)
)
(7): Conv(
(conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(8): C3k2(
(cv1): Conv(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(384, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): C3k(
(cv1): Conv(
(conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv3): Conv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(cv1): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Bottleneck(
(cv1): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
)
)
(9): SPPF(
(cv1): Conv(
(conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): MaxPool2d(kernel_size=5, stride=1, padding=2, dilation=1, ceil_mode=False)
)
(10): C2PSA(
(cv1): Conv(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): PSABlock(
(attn): Attention(
(qkv): Conv(
(conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): Identity()
)
(proj): Conv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): Identity()
)
(pe): Conv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): Identity()
)
)
(ffn): Sequential(
(0): Conv(
(conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): Identity()
)
)
)
)
)
(11): Upsample(scale_factor=2.0, mode='nearest')
(12): Concat()
(13): C3k2(
(cv1): Conv(
(conv): Conv2d(384, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(192, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): Bottleneck(
(cv1): Conv(
(conv): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
(14): Upsample(scale_factor=2.0, mode='nearest')
(15): Concat()
(16): C3k2(
(cv1): Conv(
(conv): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(96, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): Bottleneck(
(cv1): Conv(
(conv): Conv2d(32, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(16, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
(17): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(18): Concat()
(19): C3k2(
(cv1): Conv(
(conv): Conv2d(192, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(192, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): Bottleneck(
(cv1): Conv(
(conv): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(32, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
(20): Conv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(21): Concat()
(22): C3k2(
(cv1): Conv(
(conv): Conv2d(384, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(384, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): ModuleList(
(0): C3k(
(cv1): Conv(
(conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv3): Conv(
(conv): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(m): Sequential(
(0): Bottleneck(
(cv1): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Bottleneck(
(cv1): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(cv2): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
)
)
)
)
(23): Detect(
(cv2): ModuleList(
(0): Sequential(
(0): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(2): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1))
)
(1): Sequential(
(0): Conv(
(conv): Conv2d(128, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(2): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1))
)
(2): Sequential(
(0): Conv(
(conv): Conv2d(256, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(2): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1))
)
)
(cv3): ModuleList(
(0): Sequential(
(0): Sequential(
(0): DWConv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Sequential(
(0): DWConv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(2): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))
)
(1): Sequential(
(0): Sequential(
(0): DWConv(
(conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(bn): BatchNorm2d(128, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(128, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Sequential(
(0): DWConv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(2): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))
)
(2): Sequential(
(0): Sequential(
(0): DWConv(
(conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256, bias=False)
(bn): BatchNorm2d(256, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(1): Sequential(
(0): DWConv(
(conv): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
(1): Conv(
(conv): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(64, eps=0.001, momentum=0.03, affine=True, track_running_stats=True)
(act): SiLU(inplace=True)
)
)
(2): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))
)
)
(dfl): DFL(
(conv): Conv2d(16, 1, kernel_size=(1, 1), stride=(1, 1), bias=False)
)
)
)
)' does not exist

[glenn-jocher]

Thanks for the thorough trace — the failure is from calling YOLO(m) where m is a DetectionModel (a PyTorch module). The YOLO constructor accepts a path or an Ultralytics Model, so it interprets the module’s string repr as a filepath and raises FileNotFoundError; use the checkpoint path saved each epoch instead, which also preserves your custom names:

import torch
from ultralytics import YOLO

def on_fit_epoch_end(trainer):
    imgs = getattr(trainer.validator.dataloader.dataset, "im_files", [])[:1]
    if not imgs:
        return
    y = YOLO(trainer.last)  # or `trainer.best`
    with torch.inference_mode():
        results = y.predict(source=imgs, imgsz=trainer.args.imgsz, save=False, verbose=False)
        results[0].show()

last.pt is written before on_fit_epoch_end runs, so it’s available here; see save_model and the callback flow via run_callbacks. For why modules aren’t accepted by the constructor, refer to the Model class reference. If you recently upgraded and are running in Jupyter, please restart the kernel to clear any stale state.

[pzae]

It worked, this is exactly what I needed! Thank you so much for your attention and the time you spent helping me. I really appreciate your assistance and detailed explanations.

[glenn-jocher]

Great to hear it’s working—credit goes to the YOLO community and the Ultralytics team. If you want those epoch visuals auto-logged without extra code, enable a logger like W&B or ClearML; their callbacks capture validation images/metrics each epoch—see the W&B callback reference and examples in the ClearML callback reference: W&B callbacks, ClearML callbacks.