Comfy-Org/desktop

Fork 0

mirror of https://github.com/Comfy-Org/desktop.git synced 2026-11-04 23:47:43 +00:00

[GH-ISSUE #1575] Windows new install still gets PyTorch 2.8.0+cu128 after `d5f4933` (expected 2.10.0+cu130) #465

New issue

Open

opened 2026-10-31 22:13:19 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-31 22:13:19 +00:00

Owner

Originally created by @benceruleanlu on GitHub (Jan 31, 2026).
Original GitHub issue: https://github.com/Comfy-Org/desktop/issues/1575

Summary

After commit d5f49336185a4f5400534020cfc3ad266f020a70 (“Bump PyTorch to 2.10.0”), a fresh Windows install still ended up with torch 2.8.0+cu128.

Expected

Fresh Windows NVIDIA install should use torch 2.10.0+cu130, torchaudio 2.10.0+cu130, torchvision 0.25.0+cu130 per updated compiled requirements.

Actual

Fresh Windows install produced torch 2.8.0+cu128.

Context

Commit d5f4933 updates:

assets/requirements/windows_nvidia.compiled to 2.10.0+cu130
NVIDIA_TORCH_VERSION constants
CUDA mirror to cu130

Questions / Hypotheses

Did installer fall back to unpinned requirements.txt and resolve to older cu128 wheels?
Is a cached uv/pip index or mirror overriding the compiled requirements?
Are we still using a legacy torch mirror (cu12x) in some install path?

Requested info from repro

Please add:

App version/build info
Installation logs from main.log + comfyui.log
Whether installer showed any fallback/requirements.compiled error
GPU + driver version
Value of Comfy-Desktop.UV.TorchInstallMirror if customized

┆Issue is synchronized with this Notion page by Unito

Originally created by @benceruleanlu on GitHub (Jan 31, 2026). Original GitHub issue: https://github.com/Comfy-Org/desktop/issues/1575 ### Summary After commit `d5f49336185a4f5400534020cfc3ad266f020a70` (“Bump PyTorch to 2.10.0”), a **fresh Windows install** still ended up with **torch 2.8.0+cu128**. ### Expected Fresh Windows NVIDIA install should use **torch 2.10.0+cu130**, torchaudio 2.10.0+cu130, torchvision 0.25.0+cu130 per updated compiled requirements. ### Actual Fresh Windows install produced **torch 2.8.0+cu128**. ### Context Commit `d5f4933` updates: - `assets/requirements/windows_nvidia.compiled` to `2.10.0+cu130` - `NVIDIA_TORCH_VERSION` constants - CUDA mirror to `cu130` ### Questions / Hypotheses - Did installer fall back to unpinned `requirements.txt` and resolve to older cu128 wheels? - Is a cached uv/pip index or mirror overriding the compiled requirements? - Are we still using a legacy torch mirror (cu12x) in some install path? ### Requested info from repro Please add: - App version/build info - Installation logs from `main.log` + `comfyui.log` - Whether installer showed any fallback/requirements.compiled error - GPU + driver version - Value of `Comfy-Desktop.UV.TorchInstallMirror` if customized ┆Issue is synchronized with this [Notion page](https://www.notion.so/Issue-1575-Windows-new-install-still-gets-PyTorch-2-8-0-cu128-after-d5f4933-expected-2-10-0-cu13-2f96d73d36508102865ff27c408dad73) by [Unito](https://www.unito.io)

bsakaguchi added the

bug

label

2026-10-31 22:13:19 +00:00

bsakaguchi commented

2026-10-31 22:13:20 +00:00

Author

Owner

@coderabbitai[bot] commented on GitHub (Jan 31, 2026):

📝 CodeRabbit Plan Mode

Generate an implementation plan and prompts that you can use with your favorite coding agent.

Create Plan

Examples

🔗 Related PRs

Comfy-Org/desktop#1512 - Update NVIDIA PyTorch to 2.9.1 cu130 [merged]
Comfy-Org/desktop#1513 - Update NVIDIA torch version on updates [merged]
Comfy-Org/desktop#1516 - hotfix: migrate legacy torch mirror [merged]
Comfy-Org/desktop#1565 - Bump PyTorch to 2.10.0 [merged]
Comfy-Org/desktop#1567 - Update ComfyUI core to v0.11.1 [merged]

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

@coderabbitai[bot] commented on GitHub (Jan 31, 2026):  ### 📝 CodeRabbit Plan Mode Generate an implementation plan and prompts that you can use with your favorite coding agent. - [ ]  Create Plan <details> <summary>Examples</summary> - [Example 1](https://redirect.github.com/coderabbitai/git-worktree-runner/issues/29#issuecomment-3589134556) - [Example 2](https://redirect.github.com/coderabbitai/git-worktree-runner/issues/12#issuecomment-3606665167) </details> --- <details> <summary><b>🔗 Related PRs</b></summary> Comfy-Org/desktop#1512 - Update NVIDIA PyTorch to 2.9.1 cu130 [merged] Comfy-Org/desktop#1513 - Update NVIDIA torch version on updates [merged] Comfy-Org/desktop#1516 - hotfix: migrate legacy torch mirror [merged] Comfy-Org/desktop#1565 - Bump PyTorch to 2.10.0 [merged] Comfy-Org/desktop#1567 - Update ComfyUI core to v0.11.1 [merged] </details> --- <details> <summary> 🧪 Issue enrichment is currently in open beta.</summary> You can configure auto-planning by selecting labels in the issue_enrichment configuration. To disable automatic issue enrichment, add the following to your `.coderabbit.yaml`: ```yaml issue_enrichment: auto_enrich: enabled: false ``` </details> 💬 Have feedback or questions? Drop into our [discord](https://discord.gg/coderabbit)!

bsakaguchi commented

2026-10-31 22:13:20 +00:00

Author

Owner

@dosubot[bot] commented on GitHub (Jan 31, 2026):

The compiled requirements for Windows NVIDIA installs after d5f4933 do pin torch to 2.10.0+cu130 and use the correct cu130 mirror, so a fresh install should not get torch 2.8.0+cu128 unless something overrides this path. The most likely causes are: (1) the installer fell back to manual install due to a failure with the compiled requirements, (2) a legacy or user-customized torch mirror (e.g., cu128) is set in your config, or (3) a known bug where the mirror setting isn't honored during install, causing fallback to an old or hardcoded URL even on new installs [source] [source] [source].

To diagnose, please check your installation logs (main.log, comfyui.log) for any errors or fallback messages about requirements.compiled. Also, check your torch mirror setting (Comfy-Desktop.UV.TorchInstallMirror) in your config—if it points to a cu128 or cu129 URL, update it to https://download.pytorch.org/whl/cu130. If you see fallback or mirror mismatch errors in the logs, that's a strong indicator the installer didn't use the intended compiled requirements or mirror.

If you want to manually correct the torch version, you can run (inside the app terminal or venv):

uv pip install torch==2.10.0+cu130 torchvision==0.25.0+cu130 torchaudio==2.10.0+cu130 --index-url=https://download.pytorch.org/whl/cu130

If your GPU is not supported by CUDA 13.0, you may need to use an older torch version, but for supported hardware, this should resolve the mismatch [source].

Let me know if you need help interpreting the logs or updating your config.

To reply, just mention @dosu.

Share context across your team and agents. Try Dosu.

@dosubot[bot] commented on GitHub (Jan 31, 2026):  The compiled requirements for Windows NVIDIA installs after d5f4933 do pin torch to 2.10.0+cu130 and use the correct cu130 mirror, so a fresh install should not get torch 2.8.0+cu128 unless something overrides this path. The most likely causes are: (1) the installer fell back to manual install due to a failure with the compiled requirements, (2) a legacy or user-customized torch mirror (e.g., cu128) is set in your config, or (3) a known bug where the mirror setting isn't honored during install, causing fallback to an old or hardcoded URL even on new installs [[source]](https://github.com/Comfy-Org/desktop/issues/1166) [[source]](https://github.com/Comfy-Org/desktop/issues/1517) [[source]](https://github.com/Comfy-Org/desktop/issues/1522). To diagnose, please check your installation logs (`main.log`, `comfyui.log`) for any errors or fallback messages about requirements.compiled. Also, check your torch mirror setting (`Comfy-Desktop.UV.TorchInstallMirror`) in your config—if it points to a cu128 or cu129 URL, update it to https://download.pytorch.org/whl/cu130. If you see fallback or mirror mismatch errors in the logs, that's a strong indicator the installer didn't use the intended compiled requirements or mirror. If you want to manually correct the torch version, you can run (inside the app terminal or venv): ``` uv pip install torch==2.10.0+cu130 torchvision==0.25.0+cu130 torchaudio==2.10.0+cu130 --index-url=https://download.pytorch.org/whl/cu130 ``` If your GPU is not supported by CUDA 13.0, you may need to use an older torch version, but for supported hardware, this should resolve the mismatch [[source]](https://github.com/Comfy-Org/desktop/issues/1300). Let me know if you need help interpreting the logs or updating your config.  *To reply, just mention [@dosu](https://go.dosu.dev/dosubot?utm_source=github&utm_medium=bot-comment&utm_campaign=github-comment-footer-20260415&utm_content=reply-with-mention&utm_term=Comfy-Org%2Fdesktop).* --- Share context across your team and agents. Try [Dosu](https://dosu.dev?utm_source=github&utm_medium=bot-comment&utm_campaign=github-comment-footer-20260415&utm_content=knowledge-infrastructure-tagline&utm_term=Comfy-Org%2Fdesktop). [![Leave Feedback](https://img.shields.io/badge/Leave%20Feedback-555555?style=flat)](https://app.dosu.dev/response-feedback?utm_source=github&utm_medium=bot-comment&utm_campaign=github-comment-footer-20260415&utm_content=knowledge-infrastructure-feedback&utm_term=Comfy-Org%2Fdesktop#message_id=cfbaffc7-950b-41e3-9943-46f68e76770c) [![Learn about desktop with Dosu](https://img.shields.io/badge/Learn%20about%20desktop%20with%20Dosu-2f7b3f?style=flat&logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%2CPHN2ZyB3aWR0aD0iODYiIGhlaWdodD0iODkiIHZpZXdCb3g9IjAgMCA4NiA4OSIgZmlsbD0ibm9uZSIgeG1sbnM9Imh0dHA6Ly93d3cudzMub3JnLzIwMDAvc3ZnIj48cGF0aCBkPSJNNS4yOTIzNiAxMi43OTI4TDE3Ljc1OTMgNi42ODE4OFY3Mi41NjY3TDUuMjkyMzYgODQuMDYxOFYxMi43OTI4WiIgZmlsbD0iI0I0QkI5MSIvPjxwYXRoIGQ9Ik0xOC4yNTc1IDczLjExOTZMNTkuMTMyOSA3Mi43NDhMNTEuNzAxMSA4Mi40MDk1TDI5LjAzMzggODYuMjkxTDYuMjM5NjIgODUuMTU1NEwxOC4yNTc1IDczLjExOTZaIiBmaWxsPSIjNzc4NTYxIi8%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%2BPHBhdGggZmlsbC1ydWxlPSJldmVub2RkIiBjbGlwLXJ1bGU9ImV2ZW5vZGQiIGQ9Ik00MC43MDQgMC41MTgwNjZIMTcuMDQzOVY3Ni4yMjIxSDQwLjcwNEg0Mi41ODA1SDQ3LjgwMTNDNjguNzA2NCA3Ni4yMjIxIDg1LjY1MzMgNTkuMjc1MiA4NS42NTMzIDM4LjM3MDFDODUuNjUzMyAxNy40NjUgNjguNzA2MyAwLjUxODA2NiA0Ny44MDEzIDAuNTE4MDY2SDQyLjU4MDVINDAuNzA0WiIgZmlsbD0iI0YzRjZGMSIvPjxwYXRoIGQ9Ik0xNy4wNDM5IDAuNTE4MDY2Vi02LjU3OTE5SDkuOTQ2NjlWMC41MTgwNjZIMTcuMDQzOVpNMTcuMDQzOSA3Ni4yMjIxSDkuOTQ2NjlWODMuMzE5NEgxNy4wNDM5Vjc2LjIyMjFaTTE3LjA0MzkgNy42MTUzMkg0MC43MDRWLTYuNTc5MTlIMTcuMDQzOVY3LjYxNTMyWk0yNC4xNDEyIDc2LjIyMjFWMC41MTgwNjZIOS45NDY2OVY3Ni4yMjIxSDI0LjE0MTJaTTQwLjcwNCA2OS4xMjQ5SDE3LjA0MzlWODMuMzE5NEg0MC43MDRWNjkuMTI0OVpNNDIuNTgwNSA2OS4xMjQ5SDQwLjcwNFY4My4zMTk0SDQyLjU4MDVWNjkuMTI0OVpNNDcuODAxMyA2OS4xMjQ5SDQyLjU4MDVWODMuMzE5NEg0Ny44MDEzVjY5LjEyNDlaTTc4LjU1NiAzOC4zNzAxQzc4LjU1NiA1NS4zNTU1IDY0Ljc4NjcgNjkuMTI0OSA0Ny44MDEzIDY5LjEyNDlWODMuMzE5NEM3Mi42MjYxIDgzLjMxOTQgOTIuNzUwNSA2My4xOTQ5IDkyLjc1MDUgMzguMzcwMUg3OC41NTZaTTQ3LjgwMTMgNy42MTUzMkM2NC43ODY2IDcuNjE1MzIgNzguNTU2IDIxLjM4NDcgNzguNTU2IDM4LjM3MDFIOTIuNzUwNUM5Mi43NTA1IDEzLjU0NTMgNzIuNjI2IC02LjU3OTE5IDQ3LjgwMTMgLTYuNTc5MTlWNy42MTUzMlpNNDIuNTgwNSA3LjYxNTMySDQ3LjgwMTNWLTYuNTc5MTlINDIuNTgwNVY3LjYxNTMyWk00MC43MDQgNy42MTUzMkg0Mi41ODA1Vi02LjU3OTE5SDQwLjcwNFY3LjYxNTMyWiIgZmlsbD0iYmxhY2siIG1hc2s9InVybCgjZG9zdS1kLWN1dG91dCkiLz48cGF0aCBkPSJNNjguOTIxNSAzNi4wMTM1QzY4LjkyMTUgMzYuMDEzNSA2NS43MzY5IDQ5LjQ3MzggNTEuNDIzMSA0OS40NzM4QzM3LjEwOTMgNDkuNDczOCAzMi41Nzg3IDM3LjM1OTYgMzIuNTc4NyAzNi4wMTM1IiBzdHJva2U9ImJsYWNrIiBzdHJva2Utd2lkdGg9IjcuNjkxNjEiIHN0cm9rZS1saW5lY2FwPSJyb3VuZCIgc3Ryb2tlLWxpbmVqb2luPSJyb3VuZCIvPjxwYXRoIGQ9Ik0wLjM0ODYzMyA4NS40OTQ2QzAuMzQ4NjMzIDg1LjQ5NDYgMjkuNDg1NiA4NS44MzA5IDM0LjgwOSA4NS42OThDNDQuODMzNyA4NS40NDc3IDUxLjI4NzIgODQuNDAyIDU3LjUyNjkgNzguOTcyNEM2Mi44MTI5IDc0LjM3MjcgNzUuMTM0MiA1OS42ODM2IDc1LjEzNDIgNTkuNjgzNiIgc3Ryb2tlPSJibGFjayIgc3Ryb2tlLXdpZHRoPSI2LjE2NDgyIi8%2BPC9zdmc%2B)](https://github.dosu.com/Comfy-Org/desktop?utm_source=github&utm_medium=bot-comment&utm_campaign=github-comment-footer-20260415&utm_content=knowledge-infrastructure-learn-repo&utm_term=Comfy-Org%2Fdesktop) [![Add Dosu to your team](https://img.shields.io/badge/Add%20Dosu%20to%20your%20team-1f6feb?style=flat)](https://app.dosu.dev/signup?referrer=openSource&source=github-footer&utm_source=github&utm_medium=bot-comment&utm_campaign=github-comment-footer-20260415&utm_content=knowledge-infrastructure-add-team&utm_term=Comfy-Org%2Fdesktop)

bsakaguchi referenced this issue

2026-10-31 22:17:45 +00:00

[PR #465] [MERGED] Enh: add more server side arguments #887

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

Comfy-Org/desktop#465

No description provided.

Rows
Columns

[GH-ISSUE #1575] Windows new install still gets PyTorch 2.8.0+cu128 after d5f4933 (expected 2.10.0+cu130) #465

Summary

Expected

Actual

Context

Questions / Hypotheses

Requested info from repro

📝 CodeRabbit Plan Mode

[GH-ISSUE #1575] Windows new install still gets PyTorch 2.8.0+cu128 after `d5f4933` (expected 2.10.0+cu130) #465