Genevieve Buckley GenevieveBuckley

Repo: https://github.com/GenevieveBuckley/test-asv/tree/main

asv docs: https://asv.readthedocs.io/en/latest/

Solved: Previously I had error messages saying the libmamba module could not be found. Installing conda-build into the same environment you are running asv from fixed that particular problem. Sometimes I have also seen an error saying libmambapy must be less than version 2.0.0 (can downgrade to conda install libmambapy==1.5.12)

Problem 2: asv is still somehow using the defaults channel (blocked by our org)

micro-sam speed tests

8th September 2023. Results from Constantin's development/benchmarks.py script.

Summary: something about AMG is really killing the performance for MPS backends.

We'll need to do some line profiling on that part of the code to get more information.
I am suspicious that the fallback to CPU might somehow involve a lot of transferring large tensors back and forth between the cpu and mps, and perhaps that is a large part of why it is so much slower than the cpu only computation.
```
Running benchmark_amg ...
```

[W MPSFallback.mm:11] Warning: The operator 'torchvision::nms' is not currently supported on the MPS backend and will fall back to run on the CPU. This may have performance implications. (function operator())

	Scan, chunks processed: 100%\|█████████████████████\| 6/6 [00:00<00:00, 6.69it/s]
	Traceback (most recent call last):
	File "/home/supervisor/miniforge3/envs/napari-bootstrapper/lib/python3.10/site-packages/superqt/utils/_qthreading.py", line 613, in reraise
	raise e
	File "/home/supervisor/miniforge3/envs/napari-bootstrapper/lib/python3.10/site-packages/superqt/utils/_qthreading.py", line 175, in run
	result = self.work()
	File "/home/supervisor/miniforge3/envs/napari-bootstrapper/lib/python3.10/site-packages/superqt/utils/_qthreading.py", line 354, in work
	return self._func(self._args, *self._kwargs)
	File "/home/supervisor/miniforge3/envs/napari-bootstrapper/lib/python3.10/site-packages/napari_bootstrapper/_widget.py", line 331, in _run_prediction
	out_lsds = prepare_ds(

	pre-commit-hooks.nix: updating /workspace/copier repo
	pre-commit installed at .git/hooks/pre-commit
	pre-commit installed at .git/hooks/commit-msg
	direnv: No virtual environment exists. Executing `poetry install` to create one.
	Creating virtualenv copier-tqPO6HKK-py3.10 in /workspace/.pyenv_mirror/poetry/virtualenvs
	Installing dependencies from lock file

	Package operations: 70 installs, 1 update, 0 removals

	• Installing six (1.16.0)

	Windows: open the miniforge terminal prompt

	conda info --envs
	# prints a list of the virtual environments you have
	conda create --name myenv
	# create a new virtual environment

	conda activate myenv
	# activate a virtual environment