[MRV2] Revert explicit Triton JIT warmup migration (#56654)

Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
Woosuk Kwon
2026-09-13 04:07:56 +00:00
committed by GitHub
co-authored by OpenAI Codex
parent c711f740b7
commit 2671fedfc7
6 changed files with 96 additions and 227 deletions
+1 -5
View File
@@ -4,10 +4,7 @@ import numpy as np
import torch
from vllm.sampling_params import SamplingParams
from vllm.v1.sample.ops.topk_topp_sampler import (
apply_top_k_top_p,
register_top_k_top_p_warmups,
)
from vllm.v1.sample.ops.topk_topp_sampler import apply_top_k_top_p
from vllm.v1.worker.gpu.buffer_utils import UvaBackedTensor
from vllm.v1.worker.gpu.sample.gumbel import apply_temperature
from vllm.v1.worker.gpu.sample.min_p import apply_min_p
@@ -21,7 +18,6 @@ class SamplingStates:
def __init__(self, max_num_reqs: int, vocab_size: int):
self.max_num_reqs = max_num_reqs
self.vocab_size = vocab_size
register_top_k_top_p_warmups()
self.temperature = UvaBackedTensor(max_num_reqs, dtype=torch.float32)
self.top_k = UvaBackedTensor(max_num_reqs, dtype=torch.int32)