add Ascend guide (#75)

Author: rdzhu225

README.md

@@ -30,18 +30,18 @@ In theory, vllm-plugin-FL can support all models available in vLLM, as long as n
 
 ### Setup
 
-0. Install vllm from the official [v0.13.0](https://github.com/vllm-project/vllm/tree/v0.13.0) (optional if the correct version is installed) or from the fork [vllm-FL](https://github.com/flagos-ai/vllm-FL).
+1. Install vllm from the official [v0.13.0](https://github.com/vllm-project/vllm/tree/v0.13.0) (optional if the correct version is installed) or from the fork [vllm-FL](https://github.com/flagos-ai/vllm-FL).
 
 
-1. Install vllm-plugin-FL
+2. Install vllm-plugin-FL
 
-    1.1 Clone the repository:
+    2.1 Clone the repository:
 
     ```sh
     git clone https://github.com/flagos-ai/vllm-plugin-FL
     ```
 
-    1.2 install
+    2.2 install
     ```sh
     cd vllm-plugin-FL
     pip install -r requirements.txt
@@ -50,15 +50,15 @@ In theory, vllm-plugin-FL can support all models available in vLLM, as long as n
     pip install --no-build-isolation -e .
     ```
 
-2. Install [FlagGems](https://github.com/flagos-ai/FlagGems/blob/master/docs/getting-started.md#quick-installation)
+3. Install [FlagGems](https://github.com/flagos-ai/FlagGems/blob/master/docs/getting-started.md#quick-installation)
 
-    2.1 Install Build Dependencies
+    3.1 Install Build Dependencies
 
     ```sh
     pip install -U scikit-build-core==0.11 pybind11 ninja cmake
     ```
 
-    2.2 Installation FlagGems
+    3.2 Installation FlagGems
 
     ```sh
     git clone https://github.com/flagos-ai/FlagGems
@@ -68,27 +68,27 @@ In theory, vllm-plugin-FL can support all models available in vLLM, as long as n
     pip install --no-build-isolation -e .
     ```
 
-3. Install [FlagCX](https://github.com/flagos-ai/FlagCX/blob/main/docs/getting_started.md#build-and-installation)
+4. (Optional) Install [FlagCX](https://github.com/flagos-ai/FlagCX/blob/main/docs/getting_started.md#build-and-installation)
 
-    3.1 Clone the repository:
+    4.1 Clone the repository:
     ```sh
     git clone https://github.com/flagos-ai/FlagCX.git
     cd FlagCX
     git checkout -b v0.9.0
     git submodule update --init --recursive
     ```
 
-    3.2 Build the library with different flags targeting to different platforms:
+    4.2 Build the library with different flags targeting to different platforms:
     ```sh
     make USE_NVIDIA=1
     ```
 
-    3.3 Set environment
+    4.3 Set environment
     ```sh
     export FLAGCX_PATH="$PWD"
     ```
 
-    3.4 Installation FlagCX
+    4.4 Installation FlagCX
     ```sh
     cd plugin/torch/
     FLAGCX_ADAPTOR=[xxx] pip install . --no-build-isolation
@@ -100,6 +100,26 @@ In theory, vllm-plugin-FL can support all models available in vLLM, as long as n
 
 If there are multiple plugins in the current environment, you can specify use vllm-plugin-fl via VLLM_PLUGINS='fl'.
 
+### Additional Steps for Ascend
+
+1. Install [FlagTree](https://resource.flagos.net)
+
+    ```sh
+    RES="--index-url=https://resource.flagos.net/repository/flagos-pypi-hosted/simple --trusted-host=https://resource.flagos.net"
+    python3 -m pip install flagtree==0.4.0+ascend3.2 $RES
+    ```
+
+2. Set required environment variable
+
+    ```sh
+    export TRITON_ALL_BLOCKS_PARALLEL=1
+    ```
+
+3. Enable eager execution
+
+    Ascend requires eager execution. Add `enforce_eager=True` to the `LLM` constructor or pass `--enforce-eager` on the command line.
+
+
 ### Run a Task
 
 #### Offline Batched Inference