[PR #40] [MERGED] Implement stochastic rounding (CORE-200) #97

New issue

Closed

opened 2026-10-26 22:20:22 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 22:20:22 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/comfy-kitchen/pull/40
Author: @rattus128
Created: 5/19/2026
Status: ✅ Merged
Merged: 5/21/2026
Merged by: @comfyanonymous

Base: main ← Head: prs/stochastic-rounding

📝 Commits (2)

5baff55 add stochastic rnd
1e58587 stochastic-rounding: implement eager and registryify

📊 Changes

6 files changed (+251 additions, -0 deletions)

View changed files

📝 comfy_kitchen/__init__.py (+21 -0)
📝 comfy_kitchen/backends/cuda/__init__.py (+37 -0)
📝 comfy_kitchen/backends/cuda/dlpack_bindings.cpp (+48 -0)
📝 comfy_kitchen/backends/cuda/ops/per_tensor_quantize.cu (+81 -0)
📝 comfy_kitchen/backends/eager/__init__.py (+12 -0)
📝 comfy_kitchen/backends/eager/quantization.py (+52 -0)

📄 Description

Add cuda kernels for stochastic rounding as implemented in ComfyUI. Significant speedups for async offload + fp8 + lora users.

See: https://github.com/Comfy-Org/ComfyUI/pull/13971

_{🔄 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/40 **Author:** [@rattus128](https://github.com/rattus128) **Created:** 5/19/2026 **Status:** ✅ Merged **Merged:** 5/21/2026 **Merged by:** [@comfyanonymous](https://github.com/comfyanonymous) **Base:** `main` ← **Head:** `prs/stochastic-rounding` --- ### 📝 Commits (2) - [`5baff55`](https://github.com/Comfy-Org/comfy-kitchen/commit/5baff556d761b925993b046862eb5b94b1b93469) add stochastic rnd - [`1e58587`](https://github.com/Comfy-Org/comfy-kitchen/commit/1e5858774c014fe4bafa24d5b19a51447148bc15) stochastic-rounding: implement eager and registryify ### 📊 Changes **6 files changed** (+251 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `comfy_kitchen/__init__.py` (+21 -0) 📝 `comfy_kitchen/backends/cuda/__init__.py` (+37 -0) 📝 `comfy_kitchen/backends/cuda/dlpack_bindings.cpp` (+48 -0) 📝 `comfy_kitchen/backends/cuda/ops/per_tensor_quantize.cu` (+81 -0) 📝 `comfy_kitchen/backends/eager/__init__.py` (+12 -0) 📝 `comfy_kitchen/backends/eager/quantization.py` (+52 -0) </details> ### 📄 Description Add cuda kernels for stochastic rounding as implemented in ComfyUI. Significant speedups for async offload + fp8 + lora users. See: https://github.com/Comfy-Org/ComfyUI/pull/13971 --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 22:20:22 +00:00