sglang/ GB200 perf

dsr1-fp8-1k1k-max-tpt

concurrency 1,024 · 22 days ago

cron
failed
re-run

This run failed

job failed Reconciled from the expected matrix — no benchmark output was uploaded for this (config, concurrency) pair.

Commit

Speed up DeepGEMM JIT warmup with per-PP-rank parallel compile (#26567)

whybeyoung·22 days ago
PR #26567 · Speed up DeepGEMM JIT warmup with per-PP-rank parallel compile
Run
GitHub Actions26796303493-1
Slurm job
GPUs·prefill / decode
ISL / OSL1024 / 1024
Logsno logs uploaded