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 12, 09:45 UTC
[BugFix] Fix ValueError in NewRequestData repr methods (#29392)
Signed-off-by: maang <maang_h@163.com>
maang-h committed
Nov 28, 2025 at 13:42 UTC
c7ba1f6bc762af8f231e6ee885725e7401d74578
2 files changed
+42
-2
tests/v1/core/test_output.py
new
+36
@@ -0,0 +1,36 @@
1
+# SPDX-License-Identifier: Apache-2.0
2
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
3
+import torch
4
+
5
+from vllm.v1.core.sched.output import NewRequestData
6
+
7
+
8
+def _create_new_requests_data(prompt_embeds: torch.Tensor | None) -> NewRequestData:
9
+ return NewRequestData(
10
+ req_id="test_req",
11
+ prompt_token_ids=None,
12
+ mm_features=[],
13
+ sampling_params=None,
14
+ pooling_params=None,
15
+ block_ids=([],),
16
+ num_computed_tokens=0,
17
+ lora_request=None,
18
+ prompt_embeds=prompt_embeds,
19
+ )
20
+
21
+
22
+def test_repr_with_none() -> None:
23
+ """Test repr when prompt_embeds is None."""
24
+ new_requests_data = _create_new_requests_data(None)
25
+
26
+ assert "prompt_embeds_shape=None" in repr(new_requests_data)
27
+ assert "prompt_embeds_shape=None" in new_requests_data.anon_repr()
28
+
29
+
30
+def test_repr_with_multi_element_tensor() -> None:
31
+ """Test repr when prompt_embeds is a multi-element tensor."""
32
+ prompt_embeds = torch.randn(10, 768)
33
+ new_requests_data = _create_new_requests_data(prompt_embeds)
34
+
35
+ assert "prompt_embeds_shape=torch.Size([10, 768])" in repr(new_requests_data)
36
+ assert "prompt_embeds_shape=torch.Size([10, 768])" in new_requests_data.anon_repr()
vllm/v1/core/sched/output.py
+6
-2
@@ -68,7 +68,9 @@ class NewRequestData:
68
)
69
70
def __repr__(self) -> str:
71
- prompt_embeds_shape = self.prompt_embeds.shape if self.prompt_embeds else None
71
+ prompt_embeds_shape = (
72
+ self.prompt_embeds.shape if self.prompt_embeds is not None else None
73
+ )
74
return (
75
f"NewRequestData("
76
f"req_id={self.req_id},"
@@ -88,7 +90,9 @@ class NewRequestData:
90
prompt_token_ids_len = (
91
len(self.prompt_token_ids) if self.prompt_token_ids is not None else None
92
)
91
- prompt_embeds_shape = self.prompt_embeds.shape if self.prompt_embeds else None
93
+ prompt_embeds_shape = (
94
+ self.prompt_embeds.shape if self.prompt_embeds is not None else None
95
+ )
96
return (
97
f"NewRequestData("
98
f"req_id={self.req_id},"