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
Disable "Forbid direct 'import triton'" check for `vllm/triton_utils/importing.py` in an extensible way (#19783)
Signed-off-by: Andrew Feldman <afeldman@redhat.com>
afeldman-nm committed
Jun 18, 2025 at 18:08 UTC
16c16301c8218a28eefcdec07757a782e5ddc643
1 file changed
+10
tools/check_triton_import.py
+10
@@ -14,6 +14,12 @@ ALLOWED_LINES = {
14
"from vllm.triton_utils import tl, triton",
15
}
16
17
+ALLOWED_FILES = {"vllm/triton_utils/importing.py"}
18
+
19
+
20
+def is_allowed_file(current_file: str) -> bool:
21
+ return current_file in ALLOWED_FILES
22
+
23
24
def is_forbidden_import(line: str) -> bool:
25
stripped = line.strip()
@@ -25,10 +31,14 @@ def parse_diff(diff: str) -> list[str]:
31
violations = []
32
current_file = None
33
current_lineno = None
34
+ skip_allowed_file = False
35
36
for line in diff.splitlines():
37
if line.startswith("+++ b/"):
38
current_file = line[6:]
39
+ skip_allowed_file = is_allowed_file(current_file)
40
+ elif skip_allowed_file:
41
+ continue
42
elif line.startswith("@@"):
43
match = re.search(r"\+(\d+)", line)
44
if match: