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)