support pip install InternEvo

.github/workflows/upload_to_pypi.yaml

@@ -0,0 +1,40 @@
+name: Build and Upload Package to Pypi
+
+on:
+  pull_request:
+    branches:
+      - "develop"
+
+  create:
+    tags:
+      - "*"
+
+env:
+  WORKSPACE_PREFIX: $(echo $GITHUB_WORKSPACE |cut -d '/' -f 1-4)
+  SLURM_PARTITION: llm_s
+  TWINE_USERNAME: __token__
+  TWINE_PASSWORD: ${{ secrets.PYPI_API_TOKEN }}
+
+jobs:
+  build_and_upload:
+    runs-on: [t_cluster]
+    steps:
+    - name: mask env
+      run: |
+        echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
+    - uses: actions/checkout@v3
+
+    - name: Extract version from tag
+      id: extract_version
+      run: echo "::set-output name=version::$(echo ${{ github.ref }} | sed -n 's/refs\/tags\/v\(.*\)/\1/p')"
+
+    - name: Install dependencies
+      run: |
+        pip install setuptools wheel twine
+
+    - name: Build and upload package
+      run: |
+        source /mnt/petrelfs/share_data/llm_env/env/llm-flash2.0
+        srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=internlm-ut-${GITHUB_RUN_ID}-${GITHUB_JOB} -N 1 -n 1 --gres=gpu:8 python setup.py sdist bdist_wheel
+        twine upload --verbose --repository pypi dist/*
+

doc/en/install.md

@@ -25,6 +25,13 @@ export CXX=${GCC_HOME}/bin/c++
 ```
 
 ### Environment Installation
+Install through pip command:
+```bash
+pip install InternEvo
+```
+
+Or install through source code:
+
 Clone the project `InternEvo` and its dependent submodules from the github repository, as follows:
 ```bash
 git clone git@github.com:InternLM/InternEvo.git --recurse-submodules
@@ -83,4 +90,4 @@ For the local standard image built with dockerfile or pulled, use the following
 ```bash
 docker run --gpus all -it -m 500g --cap-add=SYS_PTRACE --cap-add=IPC_LOCK --shm-size 20g --network=host --name myinternlm internlm/internlm:torch1.13.1-cuda11.7.1-flashatten1.0.5-centos7 bash
 ```
-The default directory in the container is `/InternEvo`, please start training according to the [Usage](./usage.md).
+The default directory in the container is `/InternLM`, please start training according to the [Usage](./usage.md).

doc/install.md

@@ -25,16 +25,21 @@ export CXX=${GCC_HOME}/bin/c++
 ```
 
 ### 环境安装
-将项目`internlm`及其依赖子模块，从 github 仓库中 clone 下来，命令如下：
+可以通过pip命令直接安装，命令如下：
 ```bash
-git clone git@github.com:InternLM/InternLM.git --recurse-submodules
+pip install InternEvo
+```
+
+也可以通过源码安装，将项目`InternEvo`及其依赖子模块，从 github 仓库中 clone 下来，命令如下：
+```bash
+git clone git@github.com:InternLM/InternEvo.git --recurse-submodules
 ```
 
 推荐使用 conda 构建一个 Python-3.10 的虚拟环境， 并基于`requirements/`文件安装项目所需的依赖包：
 ```bash
-conda create --name internlm-env python=3.10 -y
-conda activate internlm-env
-cd internlm
+conda create --name internevo-env python=3.10 -y
+conda activate internevo-env
+cd InternEvo
 pip install -r requirements/torch.txt
 pip install -r requirements/runtime.txt
 ```
@@ -62,7 +67,7 @@ cd ../../
 用户可以使用提供的 dockerfile 结合 docker.Makefile 来构建自己的镜像，或者也可以从 https://hub.docker.com/r/internlm/internlm 获取安装了 InternLM 运行环境的镜像。
 
 #### 镜像配置及构造
-dockerfile 的配置以及构造均通过 docker.Makefile 文件实现，在 InternLM 根目录下执行如下命令即可 build 镜像：
+dockerfile 的配置以及构造均通过 docker.Makefile 文件实现，在 InternEvo 根目录下执行如下命令即可 build 镜像：
 ``` bash
 make -f docker.Makefile BASE_OS=centos7
 ```

requirements/runtime.txt

@@ -14,4 +14,5 @@ botocore
 torch-scatter
 pyecharts
 py-libnuma
+tensorboard
 -f https://data.pyg.org/whl/torch-1.13.1+cu117.html

setup.py

@@ -0,0 +1,66 @@
+import os
+import re
+import sys
+import subprocess
+from setuptools import setup, find_packages
+from setuptools.command.install import install
+
+pwd = os.path.dirname(__file__)
+
+def readme():
+    with open(os.path.join(pwd, 'README.md'), encoding='utf-8') as f:
+        content = f.read()
+    return content
+
+def get_version():
+    with open(os.path.join(pwd, 'version.txt'), 'r') as f:
+        content = f.read()
+    return content
+
+class CustomInstall(install):
+    def run(self):
+        install.run(self)
+
+        def custom_install_step(path, command):
+            original_dir = os.getcwd()
+            os.chdir(path)
+            subprocess.check_call(command, shell=True)
+            os.chdir(original_dir)
+
+        custom_install_step('./requirements', 'pip install -r torch.txt')
+        custom_install_step('./requirements', 'pip install -r runtime.txt')
+        custom_install_step('./third_party/flash-attention', 'python setup.py install')
+        custom_install_step('./third_party/flash-attention/csrc/fused_dense_lib', 'pip install -v .')
+        custom_install_step('./third_party/flash-attention/csrc/xentropy', 'pip install -v .')
+        custom_install_step('./third_party/flash-attention/csrc/rotary', 'pip install -v .')
+        custom_install_step('./third_party/flash-attention/csrc/layer_norm', 'pip install -v .')
+        custom_install_step('./third_party/apex', 'pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./')
+
+setup(
+    name='InternEvo',
+    version=get_version(),
+    description='an open-sourced lightweight training framework aims to support model pre-training without the need for extensive dependencies',
+    long_description=readme(),
+    long_description_content_type='text/markdown',
+    packages=find_packages(),
+    setup_requires=[
+        'flash-attention',
+        'apex',
+    ],
+    dependency_links=[
+        'git+https://github.com/Dao-AILab/flash-attention.git#egg=flash-attention',
+        'git+https://github.com/NVIDIA/apex.git#egg=apex',
+    ],
+    cmdclass={
+        'install': CustomInstall,
+    },
+    classifiers=[
+        'Programming Language :: Python :: 3.8',
+        'Programming Language :: Python :: 3.9',
+        'Programming Language :: Python :: 3.10',
+        'Programming Language :: Python :: 3.11',
+        'Intended Audience :: Developers',
+        'Intended Audience :: Education',
+        'Intended Audience :: Science/Research',
+    ],
+)