EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[CPU][doc] add torch.compile param in example commands (#10349)

Browse Source
This commit is contained in:
Zaili Wang
2025-09-12 10:22:46 +08:00
committed by GitHub
parent 144ee5f37c
commit 7bc5fb0d78
1 changed files with 11 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
docs/platforms/cpu_server.md
View File

@@ -139,9 +139,10 @@ Notes:
You may need to set proper `--max-total-tokens` to avoid the out-of-memory error. You may need to set proper `--max-total-tokens` to avoid the out-of-memory error.
3. For optimizing decoding with torch.compile, please add the flag `--enable-torch-compile`. 3. For optimizing decoding with torch.compile, please add the flag `--enable-torch-compile`.
To specify the maximum batch size when using torch compile, set the flag `--torch-compile-max-bs`. To specify the maximum batch size when using `torch.compile`, set the flag `--torch-compile-max-bs`.
For example, `--enable-torch-compile --torch-compile-max-bs 4` means using torch compile and setting the For example, `--enable-torch-compile --torch-compile-max-bs 4` means using `torch.compile`
maximum batch size to 4. and setting the maximum batch size to 4. Currently the maximum applicable batch size
for optimizing with `torch.compile` is 16.
4. A warmup step is automatically triggered when the service is started. 4. A warmup step is automatically triggered when the service is started.
The server is ready when you see the log `The server is fired up and ready to roll!`. The server is ready when you see the log `The server is fired up and ready to roll!`.
@@ -184,6 +185,8 @@ python -m sglang.launch_server \
--quantization w8a8_int8 \ --quantization w8a8_int8 \
--host 0.0.0.0 \ --host 0.0.0.0 \
--mem-fraction-static 0.8 \ --mem-fraction-static 0.8 \
--enable-torch-compile \
--torch-compile-max-bs 4 \
--tp 6 --tp 6
``` ```
@@ -197,8 +200,13 @@ python -m sglang.launch_server \
--device cpu \ --device cpu \
--host 0.0.0.0 \ --host 0.0.0.0 \
--mem-fraction-static 0.8 \ --mem-fraction-static 0.8 \
--enable-torch-compile \
--torch-compile-max-bs 4 \
--tp 6 --tp 6
``` ```
Note: Please set `--torch-compile-max-bs` to the maximum desired batch size for your deployment,
which can be up to 16. The value `4` in the examples is illustrative.
Then you can test with `bench_serving` command or construct your own command or script Then you can test with `bench_serving` command or construct your own command or script
following [the benchmarking example](#benchmarking-with-requests). following [the benchmarking example](#benchmarking-with-requests).