IntelLabs/Xe-Forge

Multi-stage LLM agent pipeline for optimizing Triton kernels on Intel XPU — from analysis to autotuning.

Open All

7 open issues indexed

CORRECTNESS_ONLY_STAGES is empty — skip-speedup behavior not exposed Open

IntelLabs/Xe-Forge#10 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
The optimized FlashAttention kernels Open

IntelLabs/Xe-Forge#34 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Kernel analyzer always points to Triton templates Open

IntelLabs/Xe-Forge#37 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Ask the LLM to return a patch instead of full optimized code? Open

IntelLabs/Xe-Forge#42 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Rename xe_forge.skills to xe_forge.tools with backward-compatible CLI migration Open

IntelLabs/Xe-Forge#45 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Make xe-forge-skill benchmark support optional `--spec` with baseline-derived config fallback Open

enhancement good first issue

IntelLabs/Xe-Forge#46 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026
Security: how to report a vulnerability (private reporting disabled) Open

IntelLabs/Xe-Forge#49 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 17, 2026

Get new issues in your inbox