Disable forced NVIDIA torch upgrades

[benceruleanlu]

Stop forcing NVIDIA PyTorch upgrades.

Users should control large PyTorch downloads and avoid surprise upgrades. This change removes the legacy torch mirror migration and stops treating NVIDIA cu130 minimum versions as a required package upgrade, so validation no longer triggers an automatic update. The package update flow also skips the recommended NVIDIA torch install step, keeping upgrades user-driven. Tradeoff: users can remain on older CUDA/torch builds and may hit compatibility issues; the NVIDIA driver warning still appears.

[coderabbitai]

📝 Walkthrough

Walkthrough

Removes legacy NVIDIA PyTorch mirror migration and automatic upgrade calls: deletes migrateLegacyTorchMirrorIfNeeded, removes LEGACY_NVIDIA_TORCH_MIRROR/TorchMirrorUrl imports and special-case logic, and stops automatic ensureRecommendedNvidiaTorch invocation during package updates.

Changes

Cohort / File(s)	Summary
ComfyInstallation changes src/main-process/comfyInstallation.ts	Deleted private static migrateLegacyTorchMirrorIfNeeded(); removed LEGACY_NVIDIA_TORCH_MIRROR and TorchMirrorUrl imports; removed call to the migration from fromConfig() (-19 lines).
Virtual environment logic src/virtualEnvironment.ts	Removed import of LEGACY_NVIDIA_TORCH_MIRROR; deleted special-case branch in fixDeviceMirrorMismatch() and removed NVIDIA upgrade check in hasRequirements (-9 lines).
Installation manager update flow src/install/installationManager.ts	Removed automatic call to installation.virtualEnvironment.ensureRecommendedNvidiaTorch() from updatePackages; added comment to indicate automatic NVIDIA Torch upgrades are disabled (control flow now skips automatic upgrade).

Possibly related PRs

• hotfix: migrate legacy torch mirror #1516: Reverts or removes the migration method and related constants that #1516 originally added.
• Update NVIDIA torch version on updates #1513: Alters NVIDIA PyTorch upgrade/mirror handling; touches the same virtual environment and upgrade logic.

✨ Finishing touches

• 📝 Generate docstrings

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Copilot]

Pull request overview

This PR removes the automatic migration logic that was forcing users with legacy NVIDIA torch installations (CUDA 12.9) to upgrade to the newer CUDA 13.0 mirror. The change allows users to remain on their existing PyTorch versions instead of being automatically migrated.

Changes:

• Removed automatic torch mirror migration from cu129 to cu130 for NVIDIA installations
• Removed unused imports related to the legacy migration functionality
• Removed the migrateLegacyTorchMirrorIfNeeded method and its invocation

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/virtualEnvironment.ts	Removed LEGACY_NVIDIA_TORCH_MIRROR import and auto-migration logic from fixDeviceMirrorMismatch function
src/main-process/comfyInstallation.ts	Removed migrateLegacyTorchMirrorIfNeeded method, its invocation, and related imports

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.