A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 04:45 UTC
[TPU] Fix tpu model runner test (#19995)
Signed-off-by: Chenyaaang <chenyangli@google.com>
Chenyaaang committed
Jun 23, 2025 at 16:04 UTC
33d5e29be90834fbac91051f97f853696b2f2ec5
1 file changed
+2
tests/v1/tpu/worker/test_tpu_model_runner.py
+2
@@ -6,6 +6,7 @@ import pytest
6
from vllm.attention.layer import Attention
7
from vllm.config import (CacheConfig, ModelConfig, SchedulerConfig, VllmConfig,
8
set_current_vllm_config)
9
+from vllm.pooling_params import PoolingParams
10
from vllm.sampling_params import SamplingParams
11
from vllm.utils import GiB_bytes
12
from vllm.v1.core.kv_cache_utils import (estimate_max_model_len,
@@ -71,6 +72,7 @@ def _schedule_new_request(*req_ids: str) -> SchedulerOutput:
72
mm_hashes=[],
73
mm_positions=[],
74
sampling_params=SamplingParams(),
75
+ pooling_params=PoolingParams(),
76
block_ids=([0], ), # block_ids should be tuple[list[int]]
77
num_computed_tokens=0,
78
lora_request=None,