[PR #598] docs(architecture): add bilingual DeepEP deep-dive #653

New issue

Open

opened 2026-10-27 21:04:55 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-27 21:04:55 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepEP/pull/598
Author: @tianhao909
Created: 3/31/2026
Status: 🔄 Open

Base: main ← Head: doc-260401

📝 Commits (2)

ba97b88 docs(architecture): add English DeepEP deep-dive
7ca87d9 docs(architecture): add Chinese DeepEP deep-dive

📊 Changes

16 files changed (+2539 additions, -0 deletions)

View changed files

📝 README.md (+9 -0)
➕ docs/README.md (+53 -0)
➕ docs/en/README.md (+48 -0)
➕ docs/en/architecture.md (+183 -0)
➕ docs/en/low-latency.md (+184 -0)
➕ docs/en/math-theory.md (+246 -0)
➕ docs/en/normal-kernels.md (+229 -0)
➕ docs/en/performance-tuning.md (+145 -0)
➕ docs/en/quick-start.md (+199 -0)
➕ docs/zh/README.md (+48 -0)
➕ docs/zh/architecture.md (+182 -0)
➕ docs/zh/low-latency.md (+188 -0)
➕ docs/zh/math-theory.md (+249 -0)
➕ docs/zh/normal-kernels.md (+229 -0)
➕ docs/zh/performance-tuning.md (+152 -0)
➕ docs/zh/quick-start.md (+195 -0)

📄 Description

Summary

add a source-driven DeepEP documentation tree under docs/ with both English and Chinese versions
cover architecture, quick start, normal kernels, low-latency kernels, math intuition, and performance tuning
link the repository README to the new bilingual documentation index

Validation

bash ./format.sh
python tests/test_intranode.py (requires an NVLink-capable multi-GPU machine)
python tests/test_internode.py (requires an NVLink + RDMA + NVSHMEM multi-node cluster)
python tests/test_low_latency.py (requires RDMA + NVSHMEM and the intended serving topology)

Notes

The repository does not include a separate CONTRIBUTING.md or PR template, so this change follows the existing README guidance and the format workflow in .github/workflows/format.yml.
The runtime tests are hardware-gated and cannot be honestly exercised in a generic environment without the required topology.

_{🔄 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/598 **Author:** [@tianhao909](https://github.com/tianhao909) **Created:** 3/31/2026 **Status:** 🔄 Open **Base:** `main` ← **Head:** `doc-260401` --- ### 📝 Commits (2) - [`ba97b88`](https://github.com/deepseek-ai/DeepEP/commit/ba97b8825cc0e5e43b58ab81e3af202511eed524) docs(architecture): add English DeepEP deep-dive - [`7ca87d9`](https://github.com/deepseek-ai/DeepEP/commit/7ca87d9c90235876e682975cd3dfdef61484bed8) docs(architecture): add Chinese DeepEP deep-dive ### 📊 Changes **16 files changed** (+2539 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `README.md` (+9 -0) ➕ `docs/README.md` (+53 -0) ➕ `docs/en/README.md` (+48 -0) ➕ `docs/en/architecture.md` (+183 -0) ➕ `docs/en/low-latency.md` (+184 -0) ➕ `docs/en/math-theory.md` (+246 -0) ➕ `docs/en/normal-kernels.md` (+229 -0) ➕ `docs/en/performance-tuning.md` (+145 -0) ➕ `docs/en/quick-start.md` (+199 -0) ➕ `docs/zh/README.md` (+48 -0) ➕ `docs/zh/architecture.md` (+182 -0) ➕ `docs/zh/low-latency.md` (+188 -0) ➕ `docs/zh/math-theory.md` (+249 -0) ➕ `docs/zh/normal-kernels.md` (+229 -0) ➕ `docs/zh/performance-tuning.md` (+152 -0) ➕ `docs/zh/quick-start.md` (+195 -0) </details> ### 📄 Description ## Summary - add a source-driven DeepEP documentation tree under docs/ with both English and Chinese versions - cover architecture, quick start, normal kernels, low-latency kernels, math intuition, and performance tuning - link the repository README to the new bilingual documentation index ## Validation - [x] bash ./format.sh - [ ] python tests/test_intranode.py (requires an NVLink-capable multi-GPU machine) - [ ] python tests/test_internode.py (requires an NVLink + RDMA + NVSHMEM multi-node cluster) - [ ] python tests/test_low_latency.py (requires RDMA + NVSHMEM and the intended serving topology) ## Notes - The repository does not include a separate CONTRIBUTING.md or PR template, so this change follows the existing README guidance and the format workflow in .github/workflows/format.yml. - The runtime tests are hardware-gated and cannot be honestly exercised in a generic environment without the required topology. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-27 21:04:55 +00:00

bsakaguchi referenced this issue

2026-10-27 21:05:17 +00:00

[PR #653] [CLOSED] fix #651 testing barrier #689