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, 12:15 UTC
[easy] Fix logspam on PiecewiseBackend errors (#17138)
Signed-off-by: rzou <zou3519@gmail.com>
Richard Zou committed
May 6, 2025 at 02:46 UTC
edbf2d609e6a23a105cc109844a187d72727b014
1 file changed
+2
vllm/compilation/backends.py
+2
@@ -241,6 +241,8 @@ class PiecewiseCompileInterpreter(torch.fx.Interpreter):
241
self.graph_pool = graph_pool
242
self.vllm_config = vllm_config
243
self.vllm_backend = vllm_backend
244
+ # When True, it annoyingly dumps the torch.fx.Graph on errors.
245
+ self.extra_traceback = False
246
247
def run(self, *args):
248
fake_args = [