DeepSpeed安装失败
问题
遇到的问题:
FileExistsError: [WinError 183] 当文件已存在时,无法创建该文件。: '.\deepspeed\ops\op_builder`
FileExistsError: [WinError 183] 当文件已存在时,无法创建该文件。: '.\deepspeed\ops\csrc`
可能的解决方法如下:
STEP1:进入报错的路径DeepSpeed\deepspeed\ops
STEP2:把以下两个文件夹拖出去csrc
,op_builder
或许就可以解决问题了
楼主完整的解决过程:
遇到了额外的问题:
C:\Users\Administrator\Desktop\ChatGLM3\basic_demo\DeepSpeed> .\build_win.bat ‘python’ 不是内部或外部命令,也不是可运行的程序 或批处理文件。
解决方法如下:
STEP1:检查是否python路径正确
打开命令提示符(CMD)。
输入
python --version
或py --version
。
(base) C:\Users\Administrator>python --version Python 3.11.9
路径正确
STEP2:确保 build_win.bat
文件正确调用了 Python。
理论上这个文件夹应该存在在这个路径下:
.\DeepSpeed
cmd
输入:
(base) C:\Users\Administrator>where python
输出:C:\ProgramData\anaconda3\python.exe
打开build_win.bat
将其修改:
set DISTUTILS_USE_SDK=1
set DS_BUILD_AIO=0
set DS_BUILD_CUTLASS_OPS=0
set DS_BUILD_EVOFORMER_ATTN=0
set DS_BUILD_FP_QUANTIZER=0
set DS_BUILD_INFERENCE_CORE_OPS=0
set DS_BUILD_RAGGED_DEVICE_OPS=0
set DS_BUILD_SPARSE_ATTN=0
set DS_BUILD_TRANSFORMER_INFERENCE=0
C:\ProgramData\anaconda3\python.exe setup.py bdist_wheel
:end```
“已解决”(因为报错其他的了,乐):
PS C:\Users\Administrator\Desktop\ChatGLM3\basic_demo\DeepSpeed> .\build_win.bat
[WARNING] Torch did not find cuda available, if cross-compiling or running with cpu only you can ignore this message. Ad
ding compute capability for Pascal, Volta, and Turing (compute capabilities 6.0, 6.1, 6.2)
DS_BUILD_OPS=0
Install Ops={‘deepspeed_not_implemented’: False, ‘deepspeed_ccl_comm’: False, ‘deepspeed_shm_comm’: False, ‘cpu_adam’: F
alse, ‘fused_adam’: False}
Traceback (most recent call last):
File “C:\Users\Administrator\Desktop\ChatGLM3\basic_demo\DeepSpeed\setup.py”, line 213, in
shutil.copytree(‘.\op_builder’, ‘.\deepspeed\ops\op_builder’)
File “C:\ProgramData\anaconda3\Lib\shutil.py”, line 573, in copytree
return _copytree(entries=entries, src=src, dst=dst, symlinks=symlinks,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File “C:\ProgramData\anaconda3\Lib\shutil.py”, line 471, in _copytree
os.makedirs(dst, exist_ok=dirs_exist_ok)
File “”, line 225, in makedirs
FileExistsError: [WinError 183] 当文件已存在时,无法创建该文件。: ‘.\deepspeed\ops\op_builder’
简而言之:
FileExistsError: [WinError 183] 当文件已存在时,无法创建该文件。: ‘.\deepspeed\ops\op_builder’
回到了最初的起点
STEP3解决文件存在
STEP3.1:进入报错的路径DeepSpeed\deepspeed\ops
STEP3.2:把以下两个文件夹拖出去csrc
, op_builder
或许就可以解决问题了
成功后的显示内容:
…
adding ‘deepspeed/utils/zero_to_fp32.py’
adding ‘deepspeed-0.14.5+d89e8cdf.data/scripts/deepspeed’
adding ‘deepspeed-0.14.5+d89e8cdf.data/scripts/deepspeed.pt’
adding ‘deepspeed-0.14.5+d89e8cdf.data/scripts/ds’
adding ‘deepspeed-0.14.5+d89e8cdf.data/scripts/ds_bench’
adding ‘deepspeed-0.14.5+d89e8cdf.data/scripts/ds_elastic’
adding ‘deepspeed-0.14.5+d89e8cdf.data/scripts/ds_report’
adding ‘deepspeed-0.14.5+d89e8cdf.data/scripts/ds_ssh’
adding ‘deepspeed-0.14.5+d89e8cdf.data/scripts/dsr’
adding ‘deepspeed-0.14.5+d89e8cdf.dist-info/LICENSE’
adding ‘deepspeed-0.14.5+d89e8cdf.dist-info/METADATA’
adding ‘deepspeed-0.14.5+d89e8cdf.dist-info/WHEEL’
adding ‘deepspeed-0.14.5+d89e8cdf.dist-info/entry_points.txt’
adding ‘deepspeed-0.14.5+d89e8cdf.dist-info/top_level.txt’
adding ‘deepspeed-0.14.5+d89e8cdf.dist-info/RECORD’
removing build\bdist.win-amd64\wheel
deepspeed build time = 11.571576833724976 secs
PS C:\Users\Administrator\Desktop\ChatGLM3\basic_demo\DeepSpeed>
检验语句:
python -c "import deepspeed; print(deepspeed.__version__)"
我是输出的:
(base) C:\Users\Administrator\Desktop\ChatGLM3\basic_demo\DeepSpeed>python -c “import deepspeed; print(deepspeed.version)” [2024-06-28 17:33:40,804] [WARNING] [real_accelerator.py:162:get_accelerator] Setting accelerator to CPU. If you have GPU or other accelerator, we were unable to detect it. [2024-06-28 17:33:42,584] [INFO] [real_accelerator.py:203:get_accelerator] Setting ds_accelerator to cpu (auto detect) [2024-06-28 17:33:44,037] [WARNING] [real_accelerator.py:162:get_accelerator] Setting accelerator to CPU. If you have GPU or other accelerator, we were unable to detect it. [2024-06-28 17:33:44,037] [INFO] [real_accelerator.py:203:get_accelerator] Setting ds_accelerator to cpu (auto detect) W0628 17:33:44.904000 9536 torch\distributed\elastic\multiprocessing\redirects.py:27] NOTE: Redirects are currently not supported in Windows or MacOs. 0.14.5+d89e8cdf
GPT:
成功显示
deepspeed
的版本号表明deepspeed
已成功安装。
(楼主是在按照一个b站up教程做的时候,发现出现了这个BUG,顺便贴一下原视频原资料)
29 大模型并行训练框架DeepSpeed(下)_哔哩哔哩_bilibili