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, 22:00 UTC
[Docs] Docs update for Pipeline Parallel (#6222)
Signed-off-by: Muralidhar Andoorveedu <muralidhar.andoorveedu@centml.ai> Co-authored-by: Simon Mo <simon.mo@hey.com>
Murali Andoorveedu committed
Jul 9, 2024 at 16:24 UTC
673dd4cae9340e78dd5c05843e41c38133aa29a6
1 file changed
+15
-2
docs/source/serving/distributed_serving.rst
+15
-2
@@ -3,7 +3,7 @@
3
Distributed Inference and Serving
4
=================================
5
6
-vLLM supports distributed tensor-parallel inference and serving. Currently, we support `Megatron-LM's tensor parallel algorithm <https://arxiv.org/pdf/1909.08053.pdf>`_. We manage the distributed runtime with either `Ray <https://github.com/ray-project/ray>`_ or python native multiprocessing. Multiprocessing can be used when deploying on a single node, multi-node inferencing currently requires Ray.
6
+vLLM supports distributed tensor-parallel inference and serving. Currently, we support `Megatron-LM's tensor parallel algorithm <https://arxiv.org/pdf/1909.08053.pdf>`_. We also support pipeline parallel as a beta feature for online serving. We manage the distributed runtime with either `Ray <https://github.com/ray-project/ray>`_ or python native multiprocessing. Multiprocessing can be used when deploying on a single node, multi-node inferencing currently requires Ray.
7
8
Multiprocessing will be used by default when not running in a Ray placement group and if there are sufficient GPUs available on the same node for the configured :code:`tensor_parallel_size`, otherwise Ray will be used. This default can be overridden via the :code:`LLM` class :code:`distributed-executor-backend` argument or :code:`--distributed-executor-backend` API server argument. Set it to :code:`mp` for multiprocessing or :code:`ray` for Ray. It's not required for Ray to be installed for the multiprocessing case.
9
@@ -23,6 +23,19 @@ To run multi-GPU serving, pass in the :code:`--tensor-parallel-size` argument wh
23
$ --model facebook/opt-13b \
24
$ --tensor-parallel-size 4
25
26
+You can also additionally specify :code:`--pipeline-parallel-size` to enable pipeline parallelism. For example, to run API server on 8 GPUs with pipeline parallelism and tensor parallelism:
27
+
28
+.. code-block:: console
29
+
30
+ $ python -m vllm.entrypoints.openai.api_server \
31
+ $ --model gpt2 \
32
+ $ --tensor-parallel-size 4 \
33
+ $ --pipeline-parallel-size 2 \
34
+ $ --distributed-executor-backend ray
35
+
36
+.. note::
37
+ Pipeline parallel is a beta feature. It is only supported for online serving and the ray backend for now, as well as LLaMa and GPT2 style models.
38
+
39
To scale vLLM beyond a single machine, install and start a `Ray runtime <https://docs.ray.io/en/latest/ray-core/starting-ray.html>`_ via CLI before running vLLM:
40
41
.. code-block:: console
@@ -35,7 +48,7 @@ To scale vLLM beyond a single machine, install and start a `Ray runtime <https:/
48
$ # On worker nodes
49
$ ray start --address=<ray-head-address>
50
38
-After that, you can run inference and serving on multiple machines by launching the vLLM process on the head node by setting :code:`tensor_parallel_size` to the number of GPUs to be the total number of GPUs across all machines.
51
+After that, you can run inference and serving on multiple machines by launching the vLLM process on the head node by setting :code:`tensor_parallel_size` multiplied by :code:`pipeline_parallel_size` to the number of GPUs to be the total number of GPUs across all machines.
52
53
.. warning::
54
Please make sure you downloaded the model to all the nodes, or the model is downloaded to some distributed file system that is accessible by all nodes.