Rethink the way we integrate the plugins (GPU algorithms)

[JoQCcoz]

SmartScope is built with a single environment and the GPU dependencies may become any issue as different versions of pytroch and cuda may be needed to run certain workflows and new plugins.

We need to explore a way to decouple the dependencies. Perhaps a separate conda environment per algorithm/plugin

[tiezheng-yuan-nih]

Goals : integrate Ptolemy into Smartscope. requirements:

1. instead of plugins, source code of Ptolemy Algorithm shall be integrated into Smartscope.
2. Remove CUDA (GPU accleration) from Ptolemy model.

[tiezheng-yuan-nih]

Questions for @JoQCcoz:

1. configurations of plugin is defined in worker.py. where the plugin command is launched in smartscope for launching Ptolmey?
2. Does Smartscope directly employ the model 211026_unet_9x64_ep6.torchmodel in Smartscope?
3. In smartscope, what kind of image (namely low/med/high resolution) you use Ptolemy to process?
4. where Smartscope feed the image *.mrc to Ptolemy

[tiezheng-yuan-nih]

Smartscope directly use the models ptolemy. No modeling training is needed at current status. Therefore, CUDA issue is not important in current version of smartscope processes.
Results: I recommend to lower down priority of this issue.

[JoQCcoz]

Putting this one on hold for now