Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.
feat: upgrade github actions/workflows to `uv` (#453)
* Upgrade GitHub Actions to use uv instead of pip - Replace actions/setup-python with astral-sh/setup-uv in all workflows - Update composite actions to use uv commands (uv pip, uv build, etc.) - Add uv caching with enable-cache: true and uv cache prune --ci - Add UV_TORCH_BACKEND: cpu for PyTorch installations - Update cache action to only handle conda caching * Add activate-environment: true to all astral-sh/setup-uv actions Explicitly set activate-environment to true for clarity and consistency. * Fix merge-artifacts job: add setup-uv before using uv commands * Fix mypy: install types-requests for requests library stubs * Fix Windows tests: use uv run for pytest to ensure virtual env is used * Fix package uninstall command: remove -y flag for safer execution * Fix test command line: use sys.executable for Python calls * Fix test command line: correct noqa directive for version command * reverted to original * Remove complex caching step from CI workflow * Reorganize uv cache minimization steps in CI workflow * update placement for mimize caching * Fix comments and update default age-days for cache cleanup * Fix working directory for package installation from archive * Add missing step to minimize uv cache in CI workflow * re-run * revert changes * revert package validation setup to use pip as the standard , (feedback from borda)
B
Bhimraj Yadav committed
3a46c4cfbd07a7db9d4ed1f1df51b2753e1ccbc5
Parent: 003af74
Committed by GitHub <[email protected]>
on 1/14/2026, 11:39:31 AM