update document

Author: Yesifan

README.md

@@ -34,7 +34,7 @@ pip install -r requirements.txt
   ```
 #### Notes
 - When running `npm run dev`, pay attention to the `server.proxy` configuration in `vite.config.ts`, it should match the address of the `FastAPI` service you started.
-- If you encounter errors like *Couldn't import the plugin "https://cdn.jsdelivr.net/npm/@inlang/message-lint-rule-without-source@latest/dist/index.js"*, it may be that your network cannot access `cdn.jsdelivr.net`. Please find a proxy or acceleration node and replace it in `project.inlang/settings.json`.
+- If you encounter errors like *Couldn't import the plugin "https://xxxxxxxxx.xxx/xxxxxxxxxxxx"*, it may be that your network cannot access `cdn.jsdelivr.net`. Please find a proxy or acceleration node and replace it in `project.inlang/settings.json`.
 
 ### Unit Testing
 #### Writing Unit Tests

README.zh.md

@@ -32,7 +32,7 @@ pip install -r requirements.txt
   ```
 #### 注意事项
 - 运行 `npm run dev` 时需要注意 `vite.config.ts` 中的 `server.proxy` 配置， 应该确保与你启动的 `FastAPI` 服务地址一致。
-- 如果出现 *Couldn't import the plugin "https://cdn.jsdelivr.net/npm/@inlang/message-lint-rule-without-source@latest/dist/index.js"* 类似的错误，可能是你的网络无法访问 `cdn.jsdelivr.net`，请自行寻找代理或者加速节点并在 `project.inlang/settings.json` 中进行替换。
+- 如果出现 *Couldn't import the plugin "https://xxxxxxxxx.xxx/xxxxxxxxxxxx"* 类似的错误，可能是你的网络无法访问 `cdn.jsdelivr.net`，请自行寻找代理或者加速节点并在 `project.inlang/settings.json` 中进行替换。
 
 ### 单元测试
 #### 编写单元测试

darkit/core/web/project.inlang/settings.json

@@ -6,11 +6,11 @@
     "zh"
   ],
   "modules": [
-    "https://fastly.jsdelivr.net/npm/@inlang/message-lint-rule-empty-pattern@latest/dist/index.js",
-    "https://fastly.jsdelivr.net/npm/@inlang/message-lint-rule-missing-translation@latest/dist/index.js",
-    "https://fastly.jsdelivr.net/npm/@inlang/message-lint-rule-without-source@latest/dist/index.js",
-    "https://fastly.jsdelivr.net/npm/@inlang/plugin-message-format@latest/dist/index.js",
-    "https://fastly.jsdelivr.net/npm/@inlang/plugin-m-function-matcher@latest/dist/index.js"
+    "https://jsd.onmicrosoft.cn/npm/@inlang/message-lint-rule-empty-pattern@latest/dist/index.js",
+    "https://jsd.onmicrosoft.cn/npm/@inlang/message-lint-rule-missing-translation@latest/dist/index.js",
+    "https://jsd.onmicrosoft.cn/npm/@inlang/message-lint-rule-without-source@latest/dist/index.js",
+    "https://jsd.onmicrosoft.cn/npm/@inlang/plugin-message-format@latest/dist/index.js",
+    "https://jsd.onmicrosoft.cn/npm/@inlang/plugin-m-function-matcher@latest/dist/index.js"
   ],
   "plugin.inlang.messageFormat": {
     "pathPattern": "./messages/{languageTag}.json"

docs/2.User-guide/1.Installation-guide.md

@@ -56,7 +56,15 @@ We recommend using Anaconda as the Python environment manager.
     ```
 
 ## Possible Issues
-1. If you encounter the error `RuntimeError: Error building extension 'wkv'` and see `nvcc: not found` or similar messages in the error details.
+1. If you encounter the error `Your compiler (c++) is not compatible with the compiler Pytorch was built with for this platform`.
+
+    You need to install the corresponding compiler version used when PyTorch was compiled:
+    ```bash
+    # Install g++, according to your environment
+    conda install -c conda-forge gxx_linux-64
+    ```
+
+2. If you encounter the error `RuntimeError: Error building extension 'wkv'` and see `nvcc: not found` or similar messages in the error details.
 
     This error may be due to the incorrect nvcc path. If CUDA is not installed, install CUDA. If CUDA is installed, add the CUDA path to the environment variables.
 
@@ -65,15 +73,17 @@ We recommend using Anaconda as the Python environment manager.
     ## Do not add bin at the end of the path, as it will cause duplicate bin folders
     export CUDA_HOME=/usr/local/cuda-12.3
     ```
-2. #error -- unsupported GNU version! gcc versions later than 13 are not supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.
+
+3. #error -- unsupported GNU version! gcc versions later than 13 are not supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.
 
     This error occurs because CUDA 11.4 does not support gcc versions higher than 13. You need to downgrade the gcc version. You can install gcc 13.3.0 using conda.
     ```bash
     conda install -c conda-forge gcc_linux-64=13.3.0
     # If there is a version mismatch, try the following command
     # conda install -c conda-forge gxx_linux-64=13.3.0 --override-channels -c conda-forge
     ```
-3. fatal error: cuda_runtime.h: No such file or directory
+
+4. fatal error: cuda_runtime.h: No such file or directory
 
     This error can be resolved as follows:
     ```shell
@@ -83,21 +93,21 @@ We recommend using Anaconda as the Python environment manager.
     export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/local/cuda-12.3/include
     ```
 
-4. If you encounter the error `UnicodeDecodeError: 'gbk' codec can't decode byte 0x84 in position 116: illegal multibyte sequence`,  
+5. If you encounter the error `UnicodeDecodeError: 'gbk' codec can't decode byte 0x84 in position 116: illegal multibyte sequence`,  
     This error may be due to the open function in the read_version method in setup.py not matching the file encoding. Modify the open function to specify the correct encoding, such as utf-8:
     ```python
     def read_version():
         with open(os.path.join("darkit", "__init__.py"), encoding="utf-8") as f:
     ```
 
-5. If you encounter the error `subprocess.CalledProcessError: Command '['which', 'c++']' returned non-zero exit status 1.`,  
+6. If you encounter the error `subprocess.CalledProcessError: Command '['which', 'c++']' returned non-zero exit status 1.`,  
     This error may be due to the lack of a C++ compiler. The solution is as follows:
     ```bash
     sudo apt update
     sudo apt install build-essential
     ```
 
-6. If you choose to compile Graphviz from source and get error while installing `pip` package `nnviz`, probably the path of Graphviz installation is not specified. Please append the path to Graphviz installation in `C_INCLUDE_PATH`, `LD_FLAGS` and `LD_LIBRARY_PATH`.
+7. If you choose to compile Graphviz from source and get error while installing `pip` package `nnviz`, probably the path of Graphviz installation is not specified. Please append the path to Graphviz installation in `C_INCLUDE_PATH`, `LD_FLAGS` and `LD_LIBRARY_PATH`.
     ```bash
     export C_INCLUDE_PATH=<graphviz_install_path>/include:$C_INCLUDE_PATH
     export LD_LIBRARY_PATH=<graphviz_install_path>/lib:$LD_LIBRARY_PATH

docs/zh/2.User-guide/1.Installation-guide.md

@@ -25,9 +25,9 @@
     # 安装 cupy 用于加速计算, pygraphviz 用于可视化编辑模型
     conda install -c conda-forge cupy pygraphviz
 
-    # 安装 ninja 和 gcc 12 用于编译
-    # 基于 CPU 的环境不需要安装 ninja 和 gcc 12
-    conda install -c conda-forge ninja gcc_linux-64=12
+    # 安装 ninja 用于编译
+    # 基于 CPU 的环境不需要安装 ninja
+    conda install -c conda-forge ninja
     ```
 
 5. 将本包安装到 Python 环境中：
@@ -61,7 +61,14 @@
     ```
 
 ## 可能遇到的问题
-1. 如果在执行的过程中遇到了 `RuntimeError: Error building extension 'wkv'` 的错误，并且在错误详细信息中看到了 `nvcc: not found` 或类似的消息。
+1. 如果在执行过程中遇到了 `Your compiler (c++) is not compatible with the compiler Pytorch was built with for this platform` 的错误。
+
+    需要根据 PyTorch 编译时使用的编译器版本安装相应的编译器：
+    ```bash
+    # 安装 g++, 根据你的环境
+    conda install -c conda-forge gxx_linux-64
+    ```
+2. 如果在执行的过程中遇到了 `RuntimeError: Error building extension 'wkv'` 的错误，并且在错误详细信息中看到了 `nvcc: not found` 或类似的消息。
 
     那么这个错误的原因可能是因为没有找到正确的 nvcc 的路径。如果没有安装 CUDA，则需要安装 CUDA。如果安装了 CUDA 则需要将 CUDA 的路径添加到环境变量中。
 
@@ -70,15 +77,15 @@
     ## 不要在路径后面加 bin，否则会导致重复的 bin 文件夹
     export CUDA_HOME=/usr/local/cuda-12.3
     ```
-2. #error -- unsupported GNU version! gcc versions later than 13 are not supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.
+3. #error -- unsupported GNU version! gcc versions later than 13 are not supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.
 
     这个错误是因为 CUDA 11.4 不支持比 gcc 13 更高的版本，需要降级 gcc 版本。可以使用 conda 安装 gcc 13.3.0。
     ```bash
     conda install -c conda-forge gcc_linux-64=13.3.0
     # 如果提示版本不匹配，可以尝试使用以下命令
     # conda install -c conda-forge gxx_linux-64=13.3.0 --override-channels -c conda-forge
     ```
-3. fatal error: cuda_runtime.h: No such file or directory
+4. fatal error: cuda_runtime.h: No such file or directory
 
     这个错误可以通过以下方式解决：
     ```shell
@@ -88,19 +95,21 @@
     # 将找到的路径添加到 CPLUS_INCLUDE_PATH 环境变量中
     export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/local/cuda-12.3/include
     ```
-4. 如果在执行的过程中遇到了 `UnicodeDecodeError: 'gbk' codec can't decode byte 0x84 in position 116: illegal multibyte sequence` 的错误，  
+5. 如果在执行的过程中遇到了 `UnicodeDecodeError: 'gbk' codec can't decode byte 0x84 in position 116: illegal multibyte sequence` 的错误
+
     那么这个错误的原因可能是因为 setup.py 中 read_version 方法中的 open 函数与文件的编码不符，修改 open 函数以指定正确的编码，如 utf-8：
     ```python
     def read_version():
         with open(os.path.join("DarwinKit", "__init__.py"), encoding="utf-8") as f:
     ```
-5. 如果在执行的过程中遇到了 `subprocess.CalledProcessError: Command '['which', 'c++']' returned non-zero exit status 1.`的错误，  
+6. 如果在执行的过程中遇到了 `subprocess.CalledProcessError: Command '['which', 'c++']' returned non-zero exit status 1.`的错误
+
     那么这个错误的原因可能是因为没有安装C++编译器，解决方法如下：
     ```bash
     sudo apt update
     sudo apt install build-essential
     ``` 
-6. 如果选择从源代码编译了 Graphviz 并在安装 pip 依赖项 nnviz 时遇到了问题，可能是没有指定 Graphviz 安装的路径。请在`C_INCLUDE_PATH`，`LD_FLAGS`和`LD_LIBRARY_PATH`中添加指向 Graphviz 安装位置的路径。
+7. 如果选择从源代码编译了 Graphviz 并在安装 pip 依赖项 nnviz 时遇到了问题，可能是没有指定 Graphviz 安装的路径。请在`C_INCLUDE_PATH`，`LD_FLAGS`和`LD_LIBRARY_PATH`中添加指向 Graphviz 安装位置的路径。
     ```bash
     export C_INCLUDE_PATH=<graphviz_install_path>/include:$C_INCLUDE_PATH
     export LD_LIBRARY_PATH=<graphviz_install_path>/lib:$LD_LIBRARY_PATH

environment.yml

@@ -1,13 +1,17 @@
 name: darkit
 channels:
   - nvidia
+  - pytorch
   - conda-forge
   - defaults
 dependencies:
+  - cupy
+  - ninja
+  - pytorch
+  - pygraphviz
+  - torchvision
   - python=3.9
-  - cupy=13.3
-  - ninja=1.12
-  - cuda-toolkit=12.2
-  - cuda-nvcc=12.2
-  - cuda-compiler=12.2
-  - pygraphviz=1.11
+  - pytorch-cuda=12.1
+  - cuda-nvcc=12.1
+  - cuda-toolkit=12.1
+  - cuda-compiler=12.1

init.py

@@ -0,0 +1,9 @@
+from darkit.core.lib.options import save_models_metadata
+
+try:
+    from darkit.lm.models import Metadata as lm_metadata
+
+    save_models_metadata("lm", lm_metadata)
+    print("Updated models options for lm.")
+except Exception as e:
+    print(f"Error loading lm.models: {e}")

setup.py

@@ -4,6 +4,7 @@
 """
 
 import os
+import subprocess
 from setuptools import setup, find_packages
 from darkit.core.utils import PWA_PATH
 
@@ -18,23 +19,12 @@ def read_version():
     return "0.0.1"
 
 
-def update_models_options():
-    from darkit.core.lib.options import save_models_metadata
-
-    try:
-        from darkit.lm.models import Metadata as lm_metadata
-
-        save_models_metadata("lm", lm_metadata)
-        print("Updated models options for lm.")
-    except Exception as e:
-        print(f"Error loading lm.models: {e}")
-
-
 if not PWA_PATH.exists():
     print(f"Warning: PWA build not found at {PWA_PATH}")
     exit(1)
 
-update_models_options()
+# 执行 init.py 文件
+subprocess.run(["python", "init.py"], check=True)
 
 with open("./requirements.txt", "r", encoding="utf-8") as fh:
     install_requires = fh.read()