Remove IPEX (Intel Extension for Pytorch) support.

[simonlui]

Intel Extension for Pytorch (IPEX) was an early downstream fork of Pytorch by Intel to add XPU or Intel Arc support to the project. They started in 2020 and releases had came out all the way up to Pytorch 2.8.0 in August of 2025. And then the team announced in intel/intel-extension-for-pytorch#867 in October shortly that they would be discontinuing it in favor of upstreaming support to Pytorch itself. And indeed, today, since #6069, that has been the first recommendation for any Intel GPU users. Since #10729, the instructions for IPEX installation has been removed from the README.md and given that the repository for IPEX has been archived since March 30th, I think it's about time to remove it even though the Pytorch versions it covers are technically still supported by ComfyUI since there is no upstream support now for those Pytorch versions themselves.

Not sure if this needs some sort of process to pass since this is somewhat of a pretty big change to do on one backend but I do believe because of the above reasons I mentioned, it is appropriate. Please let me know if there is anything I would need to do or change to get this through. Also, want to tag @qiacheng to get some input and feedback about this issue.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 53226625-582a-4802-83b5-c242867bbe91

📥 Commits

Reviewing files that changed from the base of the PR and between 96f1cee and a83abdd.

📒 Files selected for processing (2)

• comfy/cli_args.py
• comfy/model_management.py

💤 Files with no reviewable changes (1)

• comfy/cli_args.py

---

📝 Walkthrough

Walkthrough

This change removes Intel Extension for PyTorch (IPEX) integration from the codebase. The --disable-ipex-optimize command-line argument is eliminated from argument parsing. The global IPEX import and the conditional ipex.optimize(...) call during model loading are deleted. Intel XPU dtype selection logic in should_use_fp16() and should_use_bf16() is simplified to directly query current torch XPU capabilities without version-based branching. Additionally, XPU cache clearing now explicitly calls torch.xpu.synchronize() before torch.xpu.empty_cache().

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately and concisely describes the main change: removing IPEX support. It is specific, clear, and directly reflects the primary objective of the changeset.
Description check	✅ Passed	The description provides comprehensive context for the removal, including the history of IPEX, reasons for discontinuation, and project rationale. It is directly relevant to the changeset.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

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

_{Review rate limit: 7/8 reviews remaining, refill in 7 minutes and 30 seconds.}

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