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 8, 21:30 UTC
[Minor] Include `StreamingInput` in inputs package (#33856)
Signed-off-by: Nick Hill <nickhill123@gmail.com>
Nick Hill committed
Feb 4, 2026 at 20:38 UTC
add9f1fbd920611c2b909fe10d9b44b59650f8b7
4 files changed
+6
-4
tests/v1/e2e/test_streaming_input.py
+1
-1
@@ -19,7 +19,7 @@ import pytest
19
import pytest_asyncio
20
21
from vllm import SamplingParams
22
-from vllm.inputs.data import StreamingInput
22
+from vllm.inputs import StreamingInput
23
from vllm.outputs import RequestOutput
24
from vllm.platforms import current_platform
25
from vllm.sampling_params import RequestOutputKind
tests/v1/streaming_input/test_async_llm_streaming.py
+1
-1
@@ -7,7 +7,7 @@ from unittest.mock import AsyncMock, MagicMock
7
8
import pytest
9
10
-from vllm.inputs.data import StreamingInput
10
+from vllm.inputs import StreamingInput
11
from vllm.outputs import RequestOutput
12
from vllm.sampling_params import RequestOutputKind, SamplingParams
13
from vllm.v1.engine.async_llm import AsyncLLM
vllm/inputs/__init__.py
+2
@@ -12,6 +12,7 @@ from .data import (
12
PromptType,
13
SingletonInputs,
14
SingletonPrompt,
15
+ StreamingInput,
16
TextPrompt,
17
TokenInputs,
18
TokensPrompt,
@@ -41,4 +42,5 @@ __all__ = [
42
"build_explicit_enc_dec_prompt",
43
"to_enc_dec_tuple_list",
44
"zip_enc_dec_prompts",
45
+ "StreamingInput",
46
]
vllm/v1/engine/async_llm.py
+2
-2
@@ -16,8 +16,7 @@ from vllm import TokensPrompt
16
from vllm.config import VllmConfig
17
from vllm.engine.arg_utils import AsyncEngineArgs
18
from vllm.engine.protocol import EngineClient
19
-from vllm.inputs import PromptType
20
-from vllm.inputs.data import StreamingInput
19
+from vllm.inputs import PromptType, StreamingInput
20
from vllm.logger import init_logger
21
from vllm.lora.request import LoRARequest
22
from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalRegistry
@@ -461,6 +460,7 @@ class AsyncLLM(EngineClient):
460
self._validate_streaming_input_sampling_params(sp)
461
else:
462
sp = sampling_params
463
+ # TODO(nick): Avoid re-validating reused sampling parameters
464
req = self.input_processor.process_inputs(
465
request_id=internal_req_id,
466
prompt=input_chunk.prompt,