[PR #208] [MERGED] Int8 attention optimizations. #221

New issue

Closed

opened 2026-10-26 22:21:44 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 22:21:44 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/comfy-kitchen/pull/208
Author: @comfyanonymous
Created: 9/27/2026
Status: ✅ Merged
Merged: 9/27/2026
Merged by: @comfyanonymous

Base: main ← Head: dev/attn_bias

📝 Commits (3)

68c768c Hip int8 attention optimizations.
100d693 int8 attention optimizations.
366eaed Missing.

📊 Changes

10 files changed (+1825 additions, -438 deletions)

View changed files

📝 comfy_kitchen/backends/cuda/sage_attention/attn_utils.cuh (+113 -1)
📝 comfy_kitchen/backends/cuda/sage_attention/qk_int_sv_i8_cuda.cuh (+664 -311)
📝 comfy_kitchen/backends/cuda/sage_attention/sage_attn_launcher.cu (+33 -4)
📝 comfy_kitchen/backends/hip/__init__.py (+30 -0)
📝 comfy_kitchen/backends/hip/dlpack_bindings.cpp (+105 -6)
➕ comfy_kitchen/backends/hip/sage_attention/dense_mask.h (+10 -0)
📝 comfy_kitchen/backends/hip/sage_attention/int8_attn.hip (+396 -59)
📝 comfy_kitchen/backends/hip/sage_attention/quant_qk_int8.hip (+78 -13)
📝 comfy_kitchen/sage_attention.py (+46 -13)
📝 tests/test_int8_attention.py (+350 -31)

📄 Description

No description provided

_{🔄 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/comfy-kitchen/pull/208 **Author:** [@comfyanonymous](https://github.com/comfyanonymous) **Created:** 9/27/2026 **Status:** ✅ Merged **Merged:** 9/27/2026 **Merged by:** [@comfyanonymous](https://github.com/comfyanonymous) **Base:** `main` ← **Head:** `dev/attn_bias` --- ### 📝 Commits (3) - [`68c768c`](https://github.com/Comfy-Org/comfy-kitchen/commit/68c768cedcdcd87349b7c13091aec73a7d950df6) Hip int8 attention optimizations. - [`100d693`](https://github.com/Comfy-Org/comfy-kitchen/commit/100d693496274b4f5e96f3355de4b334907e91b9) int8 attention optimizations. - [`366eaed`](https://github.com/Comfy-Org/comfy-kitchen/commit/366eaede39bf2149a97ddb9b5f37343e181e031c) Missing. ### 📊 Changes **10 files changed** (+1825 additions, -438 deletions) <details> <summary>View changed files</summary> 📝 `comfy_kitchen/backends/cuda/sage_attention/attn_utils.cuh` (+113 -1) 📝 `comfy_kitchen/backends/cuda/sage_attention/qk_int_sv_i8_cuda.cuh` (+664 -311) 📝 `comfy_kitchen/backends/cuda/sage_attention/sage_attn_launcher.cu` (+33 -4) 📝 `comfy_kitchen/backends/hip/__init__.py` (+30 -0) 📝 `comfy_kitchen/backends/hip/dlpack_bindings.cpp` (+105 -6) ➕ `comfy_kitchen/backends/hip/sage_attention/dense_mask.h` (+10 -0) 📝 `comfy_kitchen/backends/hip/sage_attention/int8_attn.hip` (+396 -59) 📝 `comfy_kitchen/backends/hip/sage_attention/quant_qk_int8.hip` (+78 -13) 📝 `comfy_kitchen/sage_attention.py` (+46 -13) 📝 `tests/test_int8_attention.py` (+350 -31) </details> ### 📄 Description _No description provided_ --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 22:21:44 +00:00