perf: Optimize the CLI by completely avoiding Pydantic imports from everyvoice -h.

[joanise]

PR Goal?

Get the CLI even faster, by avoiding all pydantic imports. Prompted once again by typing everyvoice new<tab> and being annoyed at the delay getting results.

According to CI this PR brings everyvoice -h down from .28s to .21s, but on GPSCC I'm seeing .65 down to .50 so I care more, and where I also really wish I could optimize further...

Fixes?

My impatience. Well, no, it doesn't fix my impatience, it only satisfies a little bit of it. 😄

Feedback sought?

I'm not sure about the name fs2/types_expensive.py. First I called it fs2/shared_types.py to parallel what we do in ev/config, but it's not shared so that makes no sense. fs2/type_definitions.py was nicely self-documenting, and I feel I'm partially undoing that, so suggestions for a better name are very welcome.

Priority?

Normal

Tests added?

Should already be covered by CI.

How to test?

time everyvoice -h before and after the PR.

Confidence?

High for the code changes, low for the file name.

Version change?

no

[joanise]

Requires EveryVoiceTTS/FastSpeech2_lightning#77

[github-actions]

CLI load time: 0:00.23
Pull Request HEAD: 8898f3bdf42311909362c68e01d69a2a0b5b8649
Imports that take more than 0.1 s:
import time: self [us] | cumulative | imported package

[codecov]

Codecov Report

Attention: Patch coverage is 75.00000% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 75.98%. Comparing base (57df25a) to head (75a65db).

❗ Current head 75a65db differs from pull request most recent head 8898f3b

Please upload reports for the commit 8898f3b to get more accurate results.

Files	Patch %	Lines
everyvoice/base_cli/checkpoint.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #429      +/-   ##
==========================================
- Coverage   76.02%   75.98%   -0.04%     
==========================================
  Files          42       42              
  Lines        2757     2757              
  Branches      455      455              
==========================================
- Hits         2096     2095       -1     
- Misses        576      577       +1     
  Partials       85       85              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[roedoejet]

What about following the convention in utils with type_definitions_heavy.py or something like that? Ideally we should use the same naming convention as the division in utils in any case.

[joanise]

What about following the convention in utils with type_definitions_heavy.py or something like that?

That sounds better, I'll take a look.

Ideally we should use the same naming convention as the division in utils in any case.

✔️💯

[joanise]

Alright, investigation done, heavy.py is for some types defs but mostly utility functions requiring expensive imports. The parallel is not perfect so I won't try to match exactly, but I'll use the word "heavy" for reasonable consistency, which means I'm going to take your first suggestion: type_definitions_heavy.py.

[joanise]

@roedoejet OK, file renamed, and rebased onto main, so this is ready to re-review and possibly merge.

[SamuelLarkin]

lgtm