Day 0 DeepSeek V4 Pro FP4 GB300 disaggregated SGLang benchmarks#1157

Merged

Oseltamivir merged 73 commits intomainfrom

dsv4-fp4-gb200-dynamo-sglang-disagg

Apr 30, 2026

Commits on Apr 25, 2026

Day 0 DeepSeek V4 Pro FP4 GB200 disaggregated SGLang benchmarks
Oseltamivir
committed
Drop unsupported backend.connector field from sglang recipes
Oseltamivir
committed
Merge branch 'main' into dsv4-fp4-gb200-dynamo-sglang-disagg
Oseltamivir
authored
Drop dynamo: version: 0.8.1 — incompatible with deepseek-v4-grace-blackwell sglang fork
Oseltamivir
committed
Add dynamo: install: false — srtctl default is install=True
Oseltamivir
committed
Pin dynamo to v1.2.0-sglang-deepseek-v4-dev.1 tag (hash 21f135f5)
Oseltamivir
committed
Force deepep-mode: low_latency to work around mxfp4+DeepEP normal-dispatch bug
Oseltamivir
committed
Drop DeepEP / DP-attn / EP — fork-only mxfp4_deepseek bug, both dispatch types broken
Oseltamivir
committed
Add moe-dense-tp-size: 1 — fix shared-experts FP8 block-quant divisibility at TP=8/16
Oseltamivir
committed
Set SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=1024 in all env blocks
Oseltamivir
committed
Switch to TP=4 single-node — match PR #75 verbatim, fix FP8 block-quant
Oseltamivir
committed
Merge branch 'main' into dsv4-fp4-gb200-dynamo-sglang-disagg
Oseltamivir
authored
Restore mi355x retry changelog entries clobbered by merge
Oseltamivir
committed

Commits on Apr 26, 2026

Commits on Apr 27, 2026

Commits on Apr 28, 2026

Commits on Apr 29, 2026

Commits on Apr 30, 2026