[Core] Ensure memory is pinned prior to async h2d copy (#45424)

Signed-off-by: Nick Hill <nickhill123@gmail.com>
This commit is contained in:
Nick Hill
2026-06-20 20:02:24 -07:00
committed by GitHub
parent 8dd1b702f2
commit 7df3d7dada
49 changed files with 254 additions and 264 deletions
-6
View File
@@ -10,7 +10,6 @@ import torch
from vllm.platforms import current_platform
from vllm.sampling_params import SamplingParams
from vllm.utils.platform_utils import is_pin_memory_available
from vllm.utils.torch_utils import make_tensor_with_pad
from vllm.v1.pool.metadata import PoolingMetadata
from vllm.v1.sample.logits_processor import LogitsProcessors
@@ -236,7 +235,6 @@ def test_sampling_metadata_in_input_batch(device: str, batch_size: int):
max_model_len=1024,
max_num_batched_tokens=1024,
device=torch.device(device),
pin_memory=is_pin_memory_available(),
vocab_size=1024,
block_sizes=[1],
kernel_block_sizes=[1],
@@ -331,7 +329,6 @@ def test_swap_states_in_input_batch(device: str, batch_size: int, swap_list: lis
max_model_len=1024,
max_num_batched_tokens=1024,
device=torch.device(device),
pin_memory=is_pin_memory_available(),
vocab_size=1024,
block_sizes=[1],
kernel_block_sizes=[1],
@@ -341,7 +338,6 @@ def test_swap_states_in_input_batch(device: str, batch_size: int, swap_list: lis
max_model_len=1024,
max_num_batched_tokens=1024,
device=torch.device(device),
pin_memory=is_pin_memory_available(),
vocab_size=1024,
block_sizes=[1],
kernel_block_sizes=[1],
@@ -410,7 +406,6 @@ def test_pooling_prompt_lens_not_aliased(device: str):
max_model_len=MAX_PROMPT_SIZE + NUM_OUTPUT_TOKENS,
max_num_batched_tokens=batch_size * (MAX_PROMPT_SIZE + NUM_OUTPUT_TOKENS),
device=torch.device(device),
pin_memory=is_pin_memory_available(),
vocab_size=VOCAB_SIZE,
block_sizes=[16],
kernel_block_sizes=[16],
@@ -459,7 +454,6 @@ def test_pooling_metadata_token_id_buffers(
max_model_len=MAX_PROMPT_SIZE + NUM_OUTPUT_TOKENS,
max_num_batched_tokens=MAX_PROMPT_SIZE + NUM_OUTPUT_TOKENS,
device=torch.device("cpu"),
pin_memory=False,
vocab_size=VOCAB_SIZE,
block_sizes=[16],
kernel_block_sizes=[16],