@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 66aaa7722d
Fork 0

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 16, 04:30 UTC

[torch.compile] remove graph logging in ci (#11110)

Signed-off-by: youkaichao <youkaichao@gmail.com>

youkaichao committed Dec 11, 2024 at 10:59 UTC 66aaa7722df3d7ef9e9bd2942cab5cd0d7473174
1 file changed +5 -3
vllm/compilation/backends.py
+5 -3
@@ -287,9 +287,11 @@ class VllmBackend:
287 graph, self.compilation_config.splitting_ops)
288
289 from torch._dynamo.utils import lazy_format_graph_code
290 - logger.debug("%s", lazy_format_graph_code("before split", self.graph))
291 - logger.debug("%s", lazy_format_graph_code("after split",
292 - self.split_gm))
290 +
291 + # depyf will hook lazy_format_graph_code and dump the graph
292 + # for debugging, no need to print the graph here
293 + lazy_format_graph_code("before split", self.graph)
294 + lazy_format_graph_code("after split", self.split_gm)
295
296 compilation_counter.num_piecewise_graphs_seen += len(
297 self.piecewise_graphs)