[PR #47] [MERGED] Refactor: isolate DeepSeek V4 Flash behind plugin registries #48

New issue

Closed

opened 2026-10-29 21:06:45 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-29 21:06:45 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/kvcache-ai/sglang/pull/47
Author: @yyj6666667
Created: 5/9/2026
Status: ✅ Merged
Merged: 5/9/2026
Merged by: @yyj6666667

Base: main ← Head: dsv4-plugin-redo

📝 Commits (1)

3e647d3 refactor(dsv4): isolate DeepSeek V4 Flash behind plugin registries

📊 Changes

43 files changed (+1303 additions, -566 deletions)

View changed files

📝 python/sglang/launch_server.py (+54 -0)
➕ python/sglang/srt/configs/_model_config_dsv4.py (+122 -0)
📝 python/sglang/srt/configs/deepseek_v4.py (+120 -62)
📝 python/sglang/srt/configs/model_config.py (+6 -104)
📝 python/sglang/srt/disaggregation/decode.py (+6 -5)
📝 python/sglang/srt/disaggregation/prefill.py (+8 -4)
📝 python/sglang/srt/disaggregation/utils.py (+5 -2)
📝 python/sglang/srt/entrypoints/openai/serving_chat.py (+3 -1)
📝 python/sglang/srt/environ.py (+26 -26)
📝 python/sglang/srt/function_call/function_call_parser.py (+0 -2)
📝 python/sglang/srt/layers/attention/compressed/compressor.py (+1 -1)
📝 python/sglang/srt/layers/attention/deepseek_v4_backend_radix.py (+4 -0)
📝 python/sglang/srt/layers/attention/nsa/utils.py (+5 -139)
➕ python/sglang/srt/layers/attention/nsa/utils_v4.py (+177 -0)
📝 python/sglang/srt/layers/moe/fused_moe_triton/fused_moe.py (+37 -73)
📝 python/sglang/srt/layers/moe/fused_moe_triton/fused_moe_triton_kernels.py (+0 -15)
📝 python/sglang/srt/layers/moe/fused_moe_triton/layer.py (+15 -54)
📝 python/sglang/srt/layers/moe/kt_ep_wrapper.py (+27 -0)
📝 python/sglang/srt/layers/moe/moe_runner/deep_gemm.py (+10 -4)
➕ python/sglang/srt/layers/moe/quant_method_registry.py (+73 -0)

...and 23 more files

📄 Description

isolate DeepSeek V4 Flash behind plugin registries

_{🔄 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/kvcache-ai/sglang/pull/47 **Author:** [@yyj6666667](https://github.com/yyj6666667) **Created:** 5/9/2026 **Status:** ✅ Merged **Merged:** 5/9/2026 **Merged by:** [@yyj6666667](https://github.com/yyj6666667) **Base:** `main` ← **Head:** `dsv4-plugin-redo` --- ### 📝 Commits (1) - [`3e647d3`](https://github.com/kvcache-ai/sglang/commit/3e647d36c0096db4731588426ea9793ad064d8fb) refactor(dsv4): isolate DeepSeek V4 Flash behind plugin registries ### 📊 Changes **43 files changed** (+1303 additions, -566 deletions) <details> <summary>View changed files</summary> 📝 `python/sglang/launch_server.py` (+54 -0) ➕ `python/sglang/srt/configs/_model_config_dsv4.py` (+122 -0) 📝 `python/sglang/srt/configs/deepseek_v4.py` (+120 -62) 📝 `python/sglang/srt/configs/model_config.py` (+6 -104) 📝 `python/sglang/srt/disaggregation/decode.py` (+6 -5) 📝 `python/sglang/srt/disaggregation/prefill.py` (+8 -4) 📝 `python/sglang/srt/disaggregation/utils.py` (+5 -2) 📝 `python/sglang/srt/entrypoints/openai/serving_chat.py` (+3 -1) 📝 `python/sglang/srt/environ.py` (+26 -26) 📝 `python/sglang/srt/function_call/function_call_parser.py` (+0 -2) 📝 `python/sglang/srt/layers/attention/compressed/compressor.py` (+1 -1) 📝 `python/sglang/srt/layers/attention/deepseek_v4_backend_radix.py` (+4 -0) 📝 `python/sglang/srt/layers/attention/nsa/utils.py` (+5 -139) ➕ `python/sglang/srt/layers/attention/nsa/utils_v4.py` (+177 -0) 📝 `python/sglang/srt/layers/moe/fused_moe_triton/fused_moe.py` (+37 -73) 📝 `python/sglang/srt/layers/moe/fused_moe_triton/fused_moe_triton_kernels.py` (+0 -15) 📝 `python/sglang/srt/layers/moe/fused_moe_triton/layer.py` (+15 -54) 📝 `python/sglang/srt/layers/moe/kt_ep_wrapper.py` (+27 -0) 📝 `python/sglang/srt/layers/moe/moe_runner/deep_gemm.py` (+10 -4) ➕ `python/sglang/srt/layers/moe/quant_method_registry.py` (+73 -0) _...and 23 more files_ </details> ### 📄 Description isolate DeepSeek V4 Flash behind plugin registries --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-29 21:06:45 +00:00

closed this issue
added the
pull-request
label

No labels

pull-request

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

kvcache-ai/sglang#48

No description provided.

Rows
Columns