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 14, 05:30 UTC
[Bugfix] Add triton.language.tensor placeholder (#25649)
Signed-off-by: Agata Dobrzyniewicz <adobrzyniewicz@habana.ai>
Agata Dobrzyniewicz committed
Sep 25, 2025 at 12:31 UTC
3c2b2ccece425d459fab1e34ed0056b8be7eae0c
2 files changed
+3
tests/test_triton_utils.py
+2
@@ -69,6 +69,8 @@ def test_triton_placeholder_language():
69
assert lang.constexpr is None
70
assert lang.dtype is None
71
assert lang.int64 is None
72
+ assert lang.int32 is None
73
+ assert lang.tensor is None
74
75
76
def test_triton_placeholder_language_from_parent():
vllm/triton_utils/importing.py
+1
@@ -93,3 +93,4 @@ class TritonLanguagePlaceholder(types.ModuleType):
93
self.dtype = None
94
self.int64 = None
95
self.int32 = None
96
+ self.tensor = None