[PR #18] [MERGED] Bump all vendor bundles to torch 2.12.x #20

New issue

Closed

opened 2026-10-26 22:23:01 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 22:23:01 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/ComfyUI-Standalone-Environments/pull/18
Author: @Kosinkadink
Created: 7/28/2026
Status: ✅ Merged
Merged: 7/28/2026
Merged by: @Kosinkadink

Base: main ← Head: chore/bump-torch-2.11-remaining-vendors

📝 Commits (2)

6245044 Bump nvidia/intel/mac/cpu bundles to torch 2.11.0
01d4b8f Bump all vendor bundles to torch 2.12.x

📊 Changes

6 files changed (+16 additions, -16 deletions)

View changed files

📝 requirements-amd-win.txt (+2 -2)
📝 requirements-amd.txt (+2 -2)
📝 requirements-cpu.txt (+3 -3)
📝 requirements-intel.txt (+3 -3)
📝 requirements-mac.txt (+3 -3)
📝 requirements-nvidia.txt (+3 -3)

📄 Description

Bumps every vendor bundle to the newest stable torch minor: 2.12.1 on the pytorch.org indexes and PyPI, 2.12.0 on AMD's multi-arch index (its newest). torchvision moves to 0.27.x; torchaudio stays at 2.11.0 everywhere because upstream stopped bumping it (the mixed tuple resolves cleanly on every index).

Real-hardware validation (fresh disposable venvs, exact pinned tuples, pip check clean in every case):

cu130 2.12.1: RTX 5090, Python 3.13.3 - CUDA 13.0, arch list includes sm_120, ConvNet workload finite
xpu 2.12.1: Intel Arc B570, Python 3.13.12 - clean-room resolution of the exact bundle install command incl. the full oneAPI runtime family, XPU workload finite
mps 2.12.1: Apple M4 Mac mini, Python 3.14.6 (cp314; no 3.13 on host - the bundle build's own smoke test provides the cp313 proof) - MPS workload finite
rocm7.14.0 2.12.0 (Linux): Navi 33/gfx1102 - real SD 1.5 inference; known non-fatal rocSHMEM/libnuma startup noise, accepted for this release
rocm7.14.0 2.12.0 (Windows): validation on the W7800 in progress; merge waits for that verdict
cpu 2.12.1: Windows x64, Python 3.13.3 - finite workload

This supersedes the AMD 2.11.0 pins shipped in v0.28.0-env1; the plan is one full-matrix build so every platform moves to 2.12.x in a single release.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/Comfy-Org/ComfyUI-Standalone-Environments/pull/18 **Author:** [@Kosinkadink](https://github.com/Kosinkadink) **Created:** 7/28/2026 **Status:** ✅ Merged **Merged:** 7/28/2026 **Merged by:** [@Kosinkadink](https://github.com/Kosinkadink) **Base:** `main` ← **Head:** `chore/bump-torch-2.11-remaining-vendors` --- ### 📝 Commits (2) - [`6245044`](https://github.com/Comfy-Org/ComfyUI-Standalone-Environments/commit/624504410fb5495b19ddb63f782e14bc2e5a7ca7) Bump nvidia/intel/mac/cpu bundles to torch 2.11.0 - [`01d4b8f`](https://github.com/Comfy-Org/ComfyUI-Standalone-Environments/commit/01d4b8ff7fa17abaeaaa6b5b1167d3a1b8c348dd) Bump all vendor bundles to torch 2.12.x ### 📊 Changes **6 files changed** (+16 additions, -16 deletions) <details> <summary>View changed files</summary> 📝 `requirements-amd-win.txt` (+2 -2) 📝 `requirements-amd.txt` (+2 -2) 📝 `requirements-cpu.txt` (+3 -3) 📝 `requirements-intel.txt` (+3 -3) 📝 `requirements-mac.txt` (+3 -3) 📝 `requirements-nvidia.txt` (+3 -3) </details> ### 📄 Description Bumps every vendor bundle to the newest stable torch minor: **2.12.1** on the pytorch.org indexes and PyPI, **2.12.0** on AMD's multi-arch index (its newest). torchvision moves to 0.27.x; torchaudio stays at 2.11.0 everywhere because upstream stopped bumping it (the mixed tuple resolves cleanly on every index). Real-hardware validation (fresh disposable venvs, exact pinned tuples, `pip check` clean in every case): - **cu130 2.12.1**: RTX 5090, Python 3.13.3 - CUDA 13.0, arch list includes sm_120, ConvNet workload finite - **xpu 2.12.1**: Intel Arc B570, Python 3.13.12 - clean-room resolution of the exact bundle install command incl. the full oneAPI runtime family, XPU workload finite - **mps 2.12.1**: Apple M4 Mac mini, Python 3.14.6 (cp314; no 3.13 on host - the bundle build's own smoke test provides the cp313 proof) - MPS workload finite - **rocm7.14.0 2.12.0 (Linux)**: Navi 33/gfx1102 - real SD 1.5 inference; known non-fatal rocSHMEM/libnuma startup noise, accepted for this release - **rocm7.14.0 2.12.0 (Windows)**: validation on the W7800 in progress; merge waits for that verdict - **cpu 2.12.1**: Windows x64, Python 3.13.3 - finite workload This supersedes the AMD 2.11.0 pins shipped in v0.28.0-env1; the plan is one full-matrix build so every platform moves to 2.12.x in a single release. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 22:23:01 +00:00

closed this issue
added the
pull-request
label

No labels

pull-request

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/ComfyUI-Standalone-Environments#20

No description provided.

Rows
Columns