Add dependency environment check and add script to quick start the web UI

src/check_requirement.py

@@ -0,0 +1,99 @@
+# 为了解决和我一样的小白问题，在入门阶段环境上面，做了一些检查，希望对大家有帮助。
+import sys
+import subprocess
+from packaging import version
+
+RED_START = "\033[91m"
+RED_END = "\033[0m"
+YELLOW_START = "\033[93m"
+YELLOW_END = "\033[0m"
+
+# 包含模块及其所需的最小和推荐版本
+required_versions = {
+    'python': ('3.8', '3.10'),
+    'torch': ('1.13.1', '2.2.0'),
+    'transformers': ('4.37.2', '4.38.2'),
+    'datasets': ('2.14.3', '2.17.1'),
+    'accelerate': ('0.27.2', '0.27.2'),
+    'peft': ('0.9.0', '0.9.0'),
+    'trl': ('0.7.11', '0.7.11'),
+}
+
+
+def check_torch_cuda_versions():
+    # 执行命令行指令
+    torch_version = subprocess.run(["python", "-c", "import torch; print(torch.__version__)", ], capture_output=True,
+                                   text=True).stdout
+    cuda_version = subprocess.run(["nvidia-smi", "--query-gpu=driver_version", "--format=csv,noheader"],
+                                  capture_output=True, text=True).stdout
+
+    if "Torch" in torch_version and "CUDA" in cuda_version:
+        if torch_version == cuda_version:
+            print("Torch and CUDA versions match")
+        else:
+            print("Torch and CUDA versions may not match. Check https://pytorch.org/get-started/previous-versions/")
+    else:
+        print("Torch or CUDA version information not found")
+
+
+# 调用函数
+check_torch_cuda_versions()
+
+
+def get_version(package_name):
+    try:
+        if package_name == 'python':
+            return f"{sys.version_info.major}.{sys.version_info.minor}.{sys.version_info.micro}"
+        else:
+            module = __import__(package_name)
+            return module.__version__
+    except ModuleNotFoundError:
+        return None
+
+
+def check_version(package_name, min_version, recommended_version):
+    actual_version = get_version(package_name)
+    if actual_version is None:
+        print(
+            f"{RED_START}Error: {package_name} is not installed. Required minimum version is {min_version}. Recommended version is {recommended_version}. Please install it.{RED_END}")
+        sys.exit(1)
+    elif version.parse(actual_version) < version.parse(min_version):
+        print(
+            f"{RED_START}Warning: {package_name} version is below the minimum required {min_version}. You have {actual_version}. It's recommended to update to at least {recommended_version}.{RED_END}")
+        # Continue execution despite the warning
+    elif version.parse(actual_version) > version.parse(recommended_version):
+        print(
+            f"{YELLOW_START}Notice: {package_name} version {actual_version} is above the recommended version {recommended_version}. You may encounter unexpected behavior.{YELLOW_END}")
+
+
+def check_torch_cuda_versions():
+    torch_version = subprocess.run(["python", "-c", "import torch; print(torch.__version__)", ], capture_output=True,
+                                   text=True).stdout
+    cuda_version = subprocess.run(["nvidia-smi", "--query-gpu=driver_version", "--format=csv,noheader"],
+                                  capture_output=True, text=True).stdout
+    # 检查 torch 和 CUDA 版本是否匹配
+    if "Torch" in torch_version and "CUDA" in cuda_version:
+        if torch_version == cuda_version:
+            print("Torch and CUDA versions match")
+        else:
+            print(
+                f"{RED_START}Torch and CUDA versions may not match. Check https://pytorch.org/get-started/previous-versions/ .{RED_END}")
+    else:
+        print(f"{RED_START}Warning: Torch or CUDA version information not found {RED_END}")
+
+
+def main():
+    for package, (min_version, recommended_version) in required_versions.items():
+        check_version(package, min_version, recommended_version)
+    # 可选项，检查torch和cuda版本是否匹配 不做硬性检查
+    check_torch_cuda_versions()
+    # 如果所有检查都通过，输出版本信息和满足条件的提示
+    for package in required_versions:
+        print(f"{package.capitalize()} version: {get_version('python' if package == 'python' else package)}")
+
+    print(
+        "Congratulations, your environment meets the requirements and you can start using LLaMA-Factory now. For more information, please visit https://github.com/LLaMA-Factory/LLaMA-Factory.")
+
+
+if __name__ == "__main__":
+    main()

start-web-ui.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+# 建议使用 魔搭社区 下载资源挺快的，
+export USE_MODELSCOPE_HUB=1
+# 目前好像只能用一个显卡，暂时先这样，后续再考虑
+export CUDA_VISIBLE_DEVICES=0
+    # check requirement
+python src/check_requirement.py
+python src/train_web.py > llama-factory.log 2>&1
+
+if [ $? -eq 0 ]; then
+    echo "The llama factory webUI server has been started successfully,"
+    echo "Access the address at http://0.0.0.0:7861"
+else
+    echo "start webUI failed, Please check the log llama-factory.log"
+fi
+echo "The log path is $(pwd)/llama-factory.log"