Conversation
Disables automatic PyTorch version upgrades that were forcing all NVIDIA users to download 1.5GB packages without consent. ## What Changed - Commented out `ensureRecommendedNvidiaTorch()` call in updatePackages() - Users will no longer be forced to upgrade PyTorch on app startup ## Why The automatic upgrade introduced in PR #1513 caused several issues: - Forced 1.5GB downloads without user warning or consent - Caused installation failures for users with slow/metered connections - Blocked Chinese users when packages weren't available on mirrors - No opt-out mechanism ## User Impact - NVIDIA users will keep their current PyTorch versions - No more forced package reinstalls on app startup - Users can manually upgrade PyTorch if they want new features ## Next Steps A follow-up PR will add optional user-prompted PyTorch upgrades with proper consent dialogs and preference tracking. Related: PR #1513, PR #1525 Fixes: Forced PyTorch upgrade crisis (Jan 8-14, 2026) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
dosubot
bot
added
the
size:XS
📝 WalkthroughWalkthroughA step invoking automatic NVIDIA PyTorch upgrade during package updates has been commented out in the installation manager. User-controlled PyTorch upgrades now replace the automatic upgrade mechanism, while subsequent validation steps remain intact. Changes
Possibly related PRs
✨ Finishing touches
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
benceruleanlu
left a comment
benceruleanlu
left a comment
There was a problem hiding this comment.
We would also need to stop hasRequirements() in src/virtualEnvironment.ts from returning package-upgrade based on NVIDIA torch. Also, the legacy mirror migration in src/main-process/comfyInstallation.ts still moves users from cu129 to cu130, which would be effectively a PyTorch upgrade if they reinstall their venv as it would pull from cu130's index.
|
I incorporated the changes/fixes in ba97520 |
|
Replaced by #1543 |
2 participants
Disables automatic PyTorch version upgrades by commenting out the ensureRecommendedNvidiaTorch() call in updatePackages().
Changed src/install/installationManager.ts:385 from:
await installation.virtualEnvironment.ensureRecommendedNvidiaTorch(callbacks);
To:
// Disabled forced PyTorch upgrades - users should control when large package downloads occur
// await installation.virtualEnvironment.ensureRecommendedNvidiaTorch(callbacks);
This prevents automatic 1.5GB PyTorch package downloads on app startup. NVIDIA users will keep their current PyTorch versions until they manually upgrade or a future PR adds opt-in upgrade prompts with user consent dialogs.
Tested: ESLint, TypeScript compilation, and pre-commit hooks all pass.
Related: PR #1513 (introduced the forced upgrade), PR #1525 (will add guided UX)
┆Issue is synchronized with this Notion page by Unito