evaluate_network method possibly searches for hardcoded config.yaml filename? #2733
Description
Is there an existing issue for this?
- I have searched the existing issues
Bug description
Hello,
Running through my pipeline, I've had one specific method fail despite every other method using the proper (in our case: d1.config.yaml) file. Here's my code, and here's the error:
code:
# %%capture --no-stderr evalModel
# Track / log time, start
timeStart, startCalc = datetime.now().time(), time.time()
print(f"Time began: {timeStart.strftime('%H:%M:%S')}")
deeplabcut.evaluate_network(configPath, plotting = True)
# print(configPath)
# Track / log time, end, elapsed
timeStop, stopCalc = datetime.now().time(), time.time()
print(f"Time completed: {timeStop.strftime('%H:%M:%S')}")
print(f"Time elapsed: {round((stopCalc - startCalc)/60, 2)} minutes")
# Config path prints correctly; DLC is doing something to the path in the background.
# Times:
# All Default Settings:
I've checked with print, the variable the config file name is stored in is still correct. It looks like the method is looking for "config.yaml" instead of what I have input as the config filename. This doesn't happen with any other method in my pipeline. Looking at the error, it looks like it's hardcoded to look for config.yaml instead of *config.yaml? We have an interest in preserving config.yaml for individual camera names irrespective of directory path, so just changing the name of config.yaml to match DLC convention wouldn't solve the problem given our use-case. Let me know if there are any thoughts. We're using the conda environment, which (having updated the environment) doesn't appear to be on the most recent version of DLC / gives an update prompt, so it's possible that this was also fixed in the most recent version.
Operating System
operating system: Ubuntu 24.04
DeepLabCut version
dlc version: 2.2.2, I believe?
DeepLabCut mode
single animal
Device type
gpu: Nvidia - 3060
Steps To Reproduce
In a conda environment, run:
deeplabcut.evaluate_network(configPath, plotting = True)
where configPath ends with a file that isn't exactly config.yaml (eg. d1.config.yaml)
Relevant log output
{
"name": "FileNotFoundError",
"message": "Config file at /home/happy/programming/deepLabCut/finches/dlcProject/d1/finchD1-happy-2024-09-02/config.yaml not found. Please make sure that the file exists and/or that you passed the path of the config file correctly!",
"stack": "---------------------------------------------------------------------------
FileNotFoundError Traceback (most recent call last)
Cell In[83], line 7
4 timeStart, startCalc = datetime.now().time(), time.time()
5 print(f\"Time began: {timeStart.strftime('%H:%M:%S')}\")
----> 7 deeplabcut.evaluate_network(configPath, plotting = True)
9 # Track / log time, end, elapsed
10 timeStop, stopCalc = datetime.now().time(), time.time()
File ~/Downloads/enter/envs/finchenv/lib/python3.10/site-packages/deeplabcut/compat.py:476, in evaluate_network(config, Shuffles, trainingsetindex, plotting, show_errors, comparisonbodyparts, gputouse, rescale, modelprefix, per_keypoint_evaluation, snapshots_to_evaluate, engine, **torch_kwargs)
474 from deeplabcut.pose_estimation_pytorch.apis import evaluate_network
475 _update_device(gputouse, torch_kwargs)
--> 476 return evaluate_network(
477 config,
478 shuffles=Shuffles,
479 trainingsetindex=trainingsetindex,
480 plotting=plotting,
481 show_errors=show_errors,
482 modelprefix=modelprefix,
483 **torch_kwargs,
484 )
486 raise NotImplementedError(f\"This function is not implemented for {engine}\")
File ~/Downloads/enter/envs/finchenv/lib/python3.10/site-packages/deeplabcut/pose_estimation_pytorch/apis/evaluate.py:393, in evaluate_network(config, shuffles, trainingsetindex, snapshotindex, device, plotting, show_errors, transform, modelprefix, detector_snapshot_index)
391 for train_set_index in train_set_indices:
392 for shuffle in shuffles:
--> 393 loader = DLCLoader(
394 config=Path(cfg[\"project_path\"]) / \"config.yaml\",
395 shuffle=shuffle,
396 trainset_index=train_set_index,
397 modelprefix=modelprefix,
398 )
399 loader.evaluation_folder.mkdir(exist_ok=True, parents=True)
401 if device is not None:
File ~/Downloads/enter/envs/finchenv/lib/python3.10/site-packages/deeplabcut/pose_estimation_pytorch/data/dlcloader.py:47, in DLCLoader.__init__(self, config, trainset_index, shuffle, modelprefix)
39 \"\"\"
40 Args:
41 config: path to the DeepLabCut project config
(...)
44 modelprefix: the modelprefix for the shuffle
45 \"\"\"
46 self._project_root = Path(config).parent
---> 47 self._project_config = af.read_config(str(config))
48 self._shuffle = shuffle
49 self._train_frac = self._project_config[\"TrainingFraction\"][trainset_index]
File ~/Downloads/enter/envs/finchenv/lib/python3.10/site-packages/deeplabcut/utils/auxiliaryfunctions.py:238, in read_config(configname)
235 raise
237 else:
--> 238 raise FileNotFoundError(
239 f\"Config file at {path} not found. Please make sure that the file exists and/or that you passed the path of the config file correctly!\"
240 )
241 return cfg
FileNotFoundError: Config file at /home/happy/programming/deepLabCut/finches/dlcProject/d1/finchD1-happy-2024-09-02/config.yaml not found. Please make sure that the file exists and/or that you passed the path of the config file correctly!"
}Anything else?
No response
Code of Conduct
- I agree to follow this project's Code of Conduct