frozenleaves/verl
1
0
Fork 0
mirror of https://github.com/volcengine/verl.git synced 2025-10-20 13:43:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

docs: fix sglang installation rendering (#762)

Browse Source 
before:

![image](https://github.com/user-attachments/assets/5a7f0f23-a601-471c-b5e9-c71073c5f9d4)


after:

![image](https://github.com/user-attachments/assets/1a3ead61-296a-4ee8-bcfc-bd289c6bbfab)
This commit is contained in:
HL
2025-03-25 19:16:34 -04:00
committed by GitHub
parent 0415b27cfd
commit 6199b8c2d8
1 changed files with 25 additions and 41 deletions
Download Patch File Download Diff File Expand all files Collapse all files

66
docs/start/install.rst
View File

@ -17,32 +17,12 @@ Training backends
We recommend using **FSDP** backend to investigate, research and prototype different models, datasets and RL algorithms. The guide for using FSDP backend can be found in :doc:`FSDP Workers<../workers/fsdp_workers>`.
For users who pursue better scalability, we recommend using **Megatron-LM** backend. Currently, we support Megatron-LM v0.4 [1]_. The guide for using Megatron-LM backend can be found in :doc:`Megatron-LM Workers<../workers/megatron_workers>`.
Install verl-SGLang from scratch
-------------------------------------
**SGLang has largely support the rearch and inference workload at xAI. For verl-sglang installation, ignore the version conflicts reported by pip with vllm. And, SGLang support native API for RLHF, do not need to patch a single line of code.**
The following steps are quick installation guide for verl-SGLang.
.. code:: bash
# Create a virtual environment and use uv for quick installation
python3 -m venv ~/.python/verl-sglang && source ~/.python/verl-sglang/bin/activate
python3 -m pip install --upgrade pip && python3 -m pip install --upgrade uv
# Install verl-SGLang
git clone https://github.com/volcengine/verl verl-sglang && cd verl-sglang
python3 -m uv pip install .
# Install the latest stable version of sglang with verl support, currently, the latest version is 0.4.3.post3
# For SGLang installation, you can also refer to https://docs.sglang.ai/start/install.html
python3 -m uv pip install "sglang[all]==0.4.3.post3" --find-links https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python
For users who pursue better scalability, we recommend using **Megatron-LM** backend. Currently, we support Megatron-LM v0.11 [1]_. The guide for using Megatron-LM backend can be found in :doc:`Megatron-LM Workers<../workers/megatron_workers>`.
Install from docker image
-------------------------
We provide pre-built Docker images for quick setup.
We provide pre-built Docker images for quick setup. For SGLang usage, please follow the later sections in this doc.
Image and tag: ``whatcanyousee/verl:vemlp-th2.4.0-cu124-vllm0.6.3-ray2.10-te2.0-megatron0.11.0-v0.0.6``. See files under ``docker/`` for NGC-based image or if you want to build your own.
@ -72,29 +52,33 @@ Image and tag: ``whatcanyousee/verl:vemlp-th2.4.0-cu124-vllm0.6.3-ray2.10-te2.0-
- **Ray**: 2.10.0
- **TransformerEngine**: 2.0.0+754d2a0
Now verl has been **compatible to Megatron-LM core_r0.11.0**, and there is **no need to apply patches** to Megatron-LM. Also, the image has integrated **Megatron-LM core_r0.11.0**, located at ``/opt/nvidia/Meagtron-LM``. One more thing, because verl only use ``megatron.core`` module for now, there is **no need to modify** ``PATH`` if you have installed Megatron-LM, like this docker image.
Now verl has been **compatible to Megatron-LM core_r0.11.0**, and there is **no need to apply patches** to Megatron-LM. Also, the image has integrated **Megatron-LM core_r0.11.0**, located at ``/opt/nvidia/Meagtron-LM``. One more thing, because verl only use ``megatron.core`` module for now, there is **no need to modify** ``PATH`` if you have installed Megatron-LM with this docker image.
Install verl-SGLang from scratch
-------------------------------------
SGLang has largely support the rearch and inference workload at xAI. For verl-sglang installation, ignore the version conflicts reported by pip with vllm. And, SGLang support native API for RLHF, do not need to patch a single line of code.
The following steps are quick installation guide for verl-SGLang.
.. code:: bash
# Create a virtual environment and use uv for quick installation
python3 -m venv ~/.python/verl-sglang && source ~/.python/verl-sglang/bin/activate
python3 -m pip install --upgrade pip && python3 -m pip install --upgrade uv
# Install verl-SGLang
git clone https://github.com/volcengine/verl verl-sglang && cd verl-sglang
python3 -m uv pip install .
If you must use Megatron-LM **core_r0.4.0**, please refer to the old docker image version ``verlai/verl:vemlp-th2.4.0-cu124-vllm0.6.3-ray2.10-te1.7-v0.0.3`` in the `Docker Hub Repo: verlai/verl <https://hub.docker.com/r/verlai/verl/tags>`_, and apply the patches in the ``verl/patches`` folder.
.. code-block:: bash
cd ..
git clone -b core_v0.4.0 https://github.com/NVIDIA/Megatron-LM.git
cp verl/patches/megatron_v4.patch Megatron-LM/
cd Megatron-LM && git apply megatron_v4.patch
pip3 install -e .
export PYTHONPATH=$PYTHONPATH:$(pwd)
Or refer to patched Megatron-LM **core_r0.4.0**:
.. code-block:: bash
git clone -b core_v0.4.0_verl https://github.com/eric-haibin-lin/Megatron-LM
export PYTHONPATH=$PYTHONPATH:$(pwd)/Megatron-LM
# Install the latest stable version of sglang with verl support, currently, the latest version is 0.4.3.post3
# For SGLang installation, you can also refer to https://docs.sglang.ai/start/install.html
python3 -m uv pip install "sglang[all]==0.4.4.post1" --find-links https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python
Install from custom environment
---------------------------------
---------------------------------------------
To manage environment, we recommend using conda: