SpatiotemporalAdaptation kwargs #2282
Description
Is there an existing issue for this?
- I have searched the existing issues
Bug description
When running SpatiotemporalAdaptation adapter.adaptation_training as defined in the documentation, there is an error looking for the adapt_iterations kwargs, even though it is assigned to the adapter object in adapter.before_adapt_inference() by default as 1000.
video_path = video_path_list[0]
superanimal_name = 'superanimal_quadruped'
from deeplabcut.modelzoo.api import SpatiotemporalAdaptation
adapter = SpatiotemporalAdaptation(
video_path,
superanimal_name,
modelfolder = 'temp',
videotype = videotype,
)
adapter.before_adapt_inference()
adapter.adaptation_training() # <–––––- error
adapter.after_adapt_inference()
You can confirm by running adapter.adapt_iterations after adapter.before_adapt_inference() and seeing the value. It is fixed easily by assigning a new kwarg adapt_iterations = 1000 and feeding it into the adapter_training:
video_path = video_path_list[0]
superanimal_name = 'superanimal_quadruped'
adapt_iterations = 1000
from deeplabcut.modelzoo.api import SpatiotemporalAdaptation
adapter = SpatiotemporalAdaptation(
video_path,
superanimal_name,
modelfolder = 'temp',
videotype = videotype,
)
adapter.before_adapt_inference()
adapter.adaptation_training(adapt_iterations=adapt_iterations) # <–––––- error looking for kwargs
adapter.after_adapt_inference()
even though this step should not be needed, or else included in the specifications. Small issue, and simple solution, but worth fixing for future use.
Operating System
MacOS Ventura
DeepLabCut version
DLC 2.3.5
DeepLabCut mode
single animal
Device type
M1
Steps To Reproduce
No response
Relevant log output
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
Cell In[24], line 14
6 adapter = SpatiotemporalAdaptation(
7 video_path,
8 superanimal_name,
9 modelfolder = 'temp',
10 videotype = videotype,
11 )
13 adapter.before_adapt_inference()
---> 14 adapter.adaptation_training()
15 adapter.after_adapt_inference()
File ~/Desktop/DeepLabCut/deeplabcut/modelzoo/api/spatiotemporal_adapt.py:175, in SpatiotemporalAdaptation.adaptation_training(self, displayiters, saveiters, **kwargs)
172 if self.modelfolder != "":
173 os.makedirs(self.modelfolder, exist_ok=True)
--> 175 self.adapt_iterations = kwargs['adapt_iterations']
178 if os.path.exists(os.path.join(self.modelfolder, f"snapshot-{self.adapt_iterations}.index")):
179 print (f'model checkpoint snapshot-{self.adapt_iterations}.index exists, skipping the video adaptation')
KeyError: 'adapt_iterations'Anything else?
No response
Code of Conduct
- I agree to follow this project's Code of Conduct