@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / f964493274
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 17, 05:30 UTC

[CI] Ensure documentation build is checked in CI (#2842)

Simon Mo committed Feb 12, 2024 at 22:53 UTC f964493274c3c839b2e27453cb70f179090cd027
5 files changed +14 -1
.buildkite/test-pipeline.yaml
+7
@@ -49,3 +49,10 @@ steps:
49 commands:
50 - pip install aiohttp
51 - bash run-benchmarks.sh
52 +
53 +- label: Documentation Build
54 + working_dir: "/vllm-workspace/docs"
55 + no_gpu: True
56 + commands:
57 + - pip install -r requirements-docs.txt
58 + - SPHINXOPTS=\"-W\" make html
.buildkite/test-template.j2
+3 -1
@@ -35,13 +35,15 @@ steps:
35 - image: "{{ docker_image }}"
36 command: ["bash"]
37 args:
38 - - "-c"
38 + - '-c'
39 - "'cd {{ (step.working_dir or default_working_dir) | safe }} && {{ step.command or (step.commands | join(' && ')) | safe }}'"
40 + {% if not step.no_gpu %}
41 resources:
42 requests:
43 nvidia.com/gpu: "{{ step.num_gpus or default_num_gpu }}"
44 limits:
45 nvidia.com/gpu: "{{ step.num_gpus or default_num_gpu }}"
46 + {% endif %}
47 env:
48 - name: HF_TOKEN
49 valueFrom:
docs/source/conf.py
+2
@@ -94,3 +94,5 @@ class MockedClassDocumenter(autodoc.ClassDocumenter):
94
95
96 autodoc.ClassDocumenter = MockedClassDocumenter
97 +
98 +navigation_with_keys = False
docs/source/index.rst
+1
@@ -89,6 +89,7 @@ Documentation
89 :caption: Quantization
90
91 quantization/auto_awq
92 + quantization/fp8_e5m2_kv_cache
93
94 .. toctree::
95 :maxdepth: 2
docs/source/quantization/fp8_e5m2_kv_cache.rst
+1
@@ -9,6 +9,7 @@ The FP8 data format retains 2~3 mantissa bits and can convert float/fp16/bflaot1
9 Here is an example of how to enable this feature:
10
11 .. code-block:: python
12 +
13 from vllm import LLM, SamplingParams
14 # Sample prompts.
15 prompts = [