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 11, 03:45 UTC
Update Pytorch version update docs (#30982)
Andrey Talman committed
Dec 19, 2025 at 11:08 UTC
268a972c6280b7f7e9e67d7012164a5a8e477961
1 file changed
+9
-14
docs/contributing/ci/update_pytorch_version.md
+9
-14
@@ -77,25 +77,20 @@ This complicates the process as we cannot use the out-of-the-box
77
- `.buildkite/release-pipeline.yaml`
78
- `.buildkite/scripts/upload-wheels.sh`
79
80
-## Address long vLLM build time
80
+## Manually running vLLM builds on BuildKiteCI
81
82
-When building vLLM with a new PyTorch/CUDA version, no cache will exist
83
-in the vLLM sccache S3 bucket, causing the build job on CI to potentially take more than 5 hours
84
-and timeout. Additionally, since vLLM's fastcheck pipeline runs in read-only mode,
85
-it doesn't populate the cache, so re-running it to warm up the cache
86
-is ineffective.
82
+When building vLLM with a new PyTorch/CUDA version, the vLLM sccache S3 bucket
83
+will not have any cached artifacts, which can cause CI build jobs to exceed 5 hours.
84
+Furthermore, vLLM's fastcheck pipeline operates in read-only mode and does not
85
+populate the cache, making it ineffective for cache warm-up purposes.
86
88
-While ongoing efforts like <https://github.com/vllm-project/vllm/issues/17419>
89
-address the long build time at its source, the current workaround is to set `VLLM_CI_BRANCH`
90
-to a custom branch provided by @khluu (`VLLM_CI_BRANCH=khluu/long_build`)
91
-when manually triggering a build on Buildkite. This branch accomplishes two things:
87
+To address this, manually trigger a build on Buildkite to accomplish two objectives:
88
93
-1. Increase the timeout limit to 10 hours so that the build doesn't time out.
94
-2. Allow the compiled artifacts to be written to the vLLM sccache S3 bucket
95
-to warm it up so that future builds are faster.
89
+1. Run the complete test suite against the PyTorch RC build by setting the environment variables: `RUN_ALL=1` and `NIGHTLY=1`
90
+2. Populate the vLLM sccache S3 bucket with compiled artifacts, enabling faster subsequent builds
91
92
<p align="center" width="100%">
98
- <img width="60%" alt="Buildkite new build popup" src="https://github.com/user-attachments/assets/a8ff0fcd-76e0-4e91-b72f-014e3fdb6b94">
93
+<img width="60%" alt="Buildkite new build popup" src="https://github.com/user-attachments/assets/3b07f71b-bb18-4ca3-aeaf-da0fe79d315f" />
94
</p>
95
96
## Update all the different vLLM platforms