[6/N][warmup][DSv4] Migrate sampling, and DFlash JIT kernels (#56323)

Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: LopezCastroRoberto <rocastro@redhat.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
Roberto L. Castro
2026-09-12 17:24:24 -07:00
committed by GitHub
co-authored by Lucas Wilkinson OpenAI Codex
parent e19a3e172e
commit aed894c190
27 changed files with 2288 additions and 558 deletions
+4 -5
View File
@@ -41,9 +41,6 @@ from vllm.model_executor.warmup.qwen_vl_triton_warmup import qwen_vl_triton_warm
from vllm.model_executor.warmup.replayssm_warmup import (
replayssm_autotune_warmup,
)
from vllm.model_executor.warmup.spec_decode_rejection_warmup import (
spec_decode_rejection_warmup,
)
from vllm.model_executor.warmup.watermark_sample_warmup import (
watermark_sample_warmup,
)
@@ -178,7 +175,10 @@ def kernel_warmup(worker: "Worker", *, process_local_only: bool = False):
logger.info("JIT kernel warmup starting.")
jit_warmup_start = time.perf_counter()
try:
worker.model_runner.jit_warmup_registry.warmup()
registry = (
worker.model_runner.jit_warmup_registry # type: ignore[attr-defined]
)
registry.warmup()
except Exception:
logger.exception(
"JIT kernel warmup failed after %.2fs.",
@@ -209,7 +209,6 @@ def kernel_warmup(worker: "Worker", *, process_local_only: bool = False):
# Run next so input-prep kernels JIT against pristine runner state.
if enable_jit_warmup:
kimi_k3_triton_warmup(worker)
spec_decode_rejection_warmup(worker)
watermark_sample_warmup(worker)
qwen4_exp_qsa_triton_warmup(worker)