[PR #187] [CLOSED] Support other ways of fp8 quantization #453

New issue

Closed

opened 2026-10-26 10:00:55 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 10:00:55 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepEP/pull/187
Author: @fzyzcjy
Created: 6/4/2025
Status: ❌ Closed

Base: main ← Head: feat/fp8_quant

📝 Commits (2)

a63df64 more
8b2d648 more

📊 Changes

1 file changed (+6 additions, -1 deletions)

View changed files

📝 csrc/kernels/internode_ll.cu (+6 -1)

📄 Description

I tested the code e2e and it works. The current code is not to be merged (I just removed the old code instead of making a switch between the two versions) and I just want to hear what kind of switch do you like (compile-time flag, c++ template, bool arg, etc), and I will update it. I personally think compile-time flag is sufficient.

_{🔄 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/deepseek-ai/DeepEP/pull/187 **Author:** [@fzyzcjy](https://github.com/fzyzcjy) **Created:** 6/4/2025 **Status:** ❌ Closed **Base:** `main` ← **Head:** `feat/fp8_quant` --- ### 📝 Commits (2) - [`a63df64`](https://github.com/deepseek-ai/DeepEP/commit/a63df64aa0c45a8583d9ffb1f3cb87cea8e47572) more - [`8b2d648`](https://github.com/deepseek-ai/DeepEP/commit/8b2d648c7197fc2c80ed656f3a7037fb7b3bbbfa) more ### 📊 Changes **1 file changed** (+6 additions, -1 deletions) <details> <summary>View changed files</summary> 📝 `csrc/kernels/internode_ll.cu` (+6 -1) </details> ### 📄 Description I tested the code e2e and it works. The current code is not to be merged (I just removed the old code instead of making a switch between the two versions) and I just want to hear what kind of switch do you like (compile-time flag, c++ template, bool arg, etc), and I will update it. I personally think compile-time flag is sufficient. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 10:00:55 +00:00