Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

docker环境问题 #14

Open
zhiwenhou1227 opened this issue Dec 19, 2023 · 11 comments
Open

docker环境问题 #14

zhiwenhou1227 opened this issue Dec 19, 2023 · 11 comments

Comments

@zhiwenhou1227
Copy link

  1. 使用docker pull keyk13/ape_image:v1 拉取了在以上问题中提供的镜像
  2. 但是在容器中没有找到xformers库，pip install xformers 会安装0.0.23版本，自动更新torch版本；如果安装0.0.17版本，会有以下报错
    NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
    query : shape=(4, 1024, 16, 64) (torch.float32)
    key : shape=(4, 1024, 16, 64) (torch.float32)
    value : shape=(4, 1024, 16, 64) (torch.float32)
    attn_bias : <class 'NoneType'>
    p : 0.0
    flshattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    requires a GPU with compute capability > 7.5
    tritonflashattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    triton is not available
    requires A100 GPU
    cutlassF is not supported because:
    xFormers wasn't build with CUDA support
    smallkF is not supported because:
    xFormers wasn't build with CUDA support
    max(query.shape[-1] != value.shape[-1]) > 32
    has custom scale
    unsupported embed per head: 64

显卡v100， 不太清楚是啥问题，可以在docker 镜像中更新可用的xformers吗
@shenyunhang
Copy link
Owner

@zhiwenhou1227

这个镜像不是我们提供的。
我们用的xformers等依赖库是从源码编译的，因为cuda版本的限制，直接pip安装可能不兼容。

在pytorch1.12.1，cuda11.3，python3.9，v100的环境下，根据官方教程，我们可以正常编译和运行以下依赖库。
xformers==0.0.23
apex==22.03
bitsandbytes==0.41.0

另外，使用更高的torch版本和对应的依赖库也可以的。

@kongdebug
Copy link

  1. 使用docker pull keyk13/ape_image:v1 拉取了在以上问题中提供的镜像
  2. 但是在容器中没有找到xformers库，pip install xformers 会安装0.0.23版本，自动更新torch版本；如果安装0.0.17版本，会有以下报错
    NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
    query : shape=(4, 1024, 16, 64) (torch.float32)
    key : shape=(4, 1024, 16, 64) (torch.float32)
    value : shape=(4, 1024, 16, 64) (torch.float32)
    attn_bias : <class 'NoneType'>
    p : 0.0
    flshattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    requires a GPU with compute capability > 7.5
    tritonflashattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    triton is not available
    requires A100 GPU
    cutlassF is not supported because:
    xFormers wasn't build with CUDA support
    smallkF is not supported because:
    xFormers wasn't build with CUDA support
    max(query.shape[-1] != value.shape[-1]) > 32
    has custom scale
    unsupported embed per head: 64

显卡v100， 不太清楚是啥问题，可以在docker 镜像中更新可用的xformers吗

你好，之前docker镜像中确实没下xformers，因为作者配置文件中没写而且不需要xformers也能运行。而在该环境中安装xformers我也遇到了问题。目前我是在torch2.1.2+cu118的环境中配置好了xformers与ape的依赖，并上传了该镜像ape_cu118

@Cristhine
Copy link

  1. 使用docker pull keyk13/ape_image:v1 拉取了在以上问题中提供的镜像
  2. 但是在容器中没有找到xformers库，pip install xformers 会安装0.0.23版本，自动更新torch版本；如果安装0.0.17版本，会有以下报错
    NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
    query : shape=(4, 1024, 16, 64) (torch.float32)
    key : shape=(4, 1024, 16, 64) (torch.float32)
    value : shape=(4, 1024, 16, 64) (torch.float32)
    attn_bias : <class 'NoneType'>
    p : 0.0
    flshattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    requires a GPU with compute capability > 7.5
    tritonflashattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    triton is not available
    requires A100 GPU
    cutlassF is not supported because:
    xFormers wasn't build with CUDA support
    smallkF is not supported because:
    xFormers wasn't build with CUDA support
    max(query.shape[-1] != value.shape[-1]) > 32
    has custom scale
    unsupported embed per head: 64

显卡v100， 不太清楚是啥问题，可以在docker 镜像中更新可用的xformers吗

你好，之前docker镜像中确实没下xformers，因为作者配置文件中没写而且不需要xformers也能运行。而在该环境中安装xformers我也遇到了问题。目前我是在torch2.1.2+cu118的环境中配置好了xformers与ape的依赖，并上传了该镜像ape_cu118

你好我使用你得镜像下载使用 进去之后执行python命令 环境里面连torch都没 请问这个镜像的python哪个环境是正常的呢

@kongdebug
Copy link

kongdebug commented Jan 10, 2024 via email

@Cristhine
Copy link

环境名称ape
…
---原始邮件--- 发件人: @.> 发送时间: 2024年1月10日(周三) 下午3:33 收件人: @.>; 抄送: @.@.>; 主题: Re: [shenyunhang/APE] docker环境问题 (Issue #14) 使用docker pull keyk13/ape_image:v1 拉取了在以上问题中提供的镜像 但是在容器中没有找到xformers库，pip install xformers 会安装0.0.23版本，自动更新torch版本；如果安装0.0.17版本，会有以下报错 NotImplementedError: No operator found for memory_efficient_attention_forward with inputs: query : shape=(4, 1024, 16, 64) (torch.float32) key : shape=(4, 1024, 16, 64) (torch.float32) value : shape=(4, 1024, 16, 64) (torch.float32) attn_bias : <class 'NoneType'> p : 0.0 flshattF is not supported because: xFormers wasn't build with CUDA support dtype=torch.float32 (supported: {torch.bfloat16, torch.float16}) requires a GPU with compute capability > 7.5 tritonflashattF is not supported because: xFormers wasn't build with CUDA support dtype=torch.float32 (supported: {torch.bfloat16, torch.float16}) triton is not available requires A100 GPU cutlassF is not supported because: xFormers wasn't build with CUDA support smallkF is not supported because: xFormers wasn't build with CUDA support max(query.shape[-1] != value.shape[-1]) > 32 has custom scale unsupported embed per head: 64 显卡v100， 不太清楚是啥问题，可以在docker 镜像中更新可用的xformers吗 你好，之前docker镜像中确实没下xformers，因为作者配置文件中没写而且不需要xformers也能运行。而在该环境中安装xformers我也遇到了问题。目前我是在torch2.1.2+cu118的环境中配置好了xformers与ape的依赖，并上传了该镜像ape_cu118 你好我使用你得镜像下载使用 进去之后执行python命令 环境里面连torch都没 请问这个镜像的python哪个环境是正常的呢 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

请问您在什么显卡环境infer或者训练的 我在a100卡上跑个infer会报错 error in ms_deformable_im2col_cuda: no kernel image is available for execution on the device 看起来是架构不对

@kongdebug
Copy link

kongdebug commented Jan 10, 2024 via email

@shutu-1
Copy link

shutu-1 commented Jan 16, 2024

你试着在容器里面运行
python3 -m pip install -e .
我是这样后可以的

环境名称ape ... ---原始邮件--- 发件人： @.> 发送时间： 2024年1月10日（周三） 下午3：33 收件人： _@**._>;抄送： _@.@。_>;主题： Re： [shenyunhang/APE] docker环境问题 （Issue #14） 使用docker pull keyk13/ape_image：v1 拉取了在以上问题中提供的镜像 但是在容器中没有找到xformers库，pip install xformers 会安装0.0.23版本，自动更新torch版本;如果安装0.0.17版本，会有以下报错 NotImplementedError： No operator found for memory_efficient_attention_forward with inputs： query ： shape=（4， 1024， 16， 64） （torch.float32） 键： shape=（4， 1024， 16， 64） （torch.float32） 值： shape=（4， 1024， 16， 64） （torch.float32） attn_bias ： <class 'NoneType'> p ： 不支持 0.0 flshattF，因为： xFormers 不是在 CUDA 支持下构建的 dtype=torch.float32（支持：{torch.bfloat16， torch.float16}） 需要具有计算能力的 GPU > 7.5 不支持 tritonflashattF，因为： xFormers 不是在 CUDA 支持下构建的 dtype=torch.float32（支持：{torch.bfloat16、torch.float16}） triton 不可用 需要 A100 GPU 不支持 cutlassF，因为： xFormers 不是在 CUDA 支持下构建的 不支持 smallkF，因为： xFormers 不是在 CUDA 支持下构建的 max（query.shape[-1] ！= value.shape[-1]） > 32 具有自定义比例 不支持 每头嵌入：64 显卡v100， 不太清楚是啥问题，可以在docker 镜像中更新可用的xformers吗 你好，之前docker镜像中确实没下xformers，因为作者配置文件中没写而且不需要xformers也能运行。而在该环境中安装xformers我也遇到了问题。目前我是在torch2.1.2+cu118的环境中配置好了xformers与ape的依赖，并上传了该镜像ape_cu118 你好我使用你得镜像下载使用 进去之后执行python命令 环境里面连火炬都没 请问这个镜像的python哪个环境是正常的呢 — Reply to this email directly to be view on GitHub， or unsubscribe.您收到此消息是因为您发表了评论。邮件 ID： _@_.*>

请问您在什么显卡环境infer或者训练的 我在a100卡上跑个infer会报错 error in ms_deformable_im2col_cuda： no kernel image is available for execution on the device 看起来是架构不对

@wdc233
Copy link

wdc233 commented Jun 6, 2024

你之前没有xformer如何运行的呀，我这没有xformer压根运行不了呀

  1. 使用docker pull keyk13/ape_image:v1 拉取了在以上问题中提供的镜像
  2. 但是在容器中没有找到xformers库，pip install xformers 会安装0.0.23版本，自动更新torch版本；如果安装0.0.17版本，会有以下报错
    NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
    query : shape=(4, 1024, 16, 64) (torch.float32)
    key : shape=(4, 1024, 16, 64) (torch.float32)
    value : shape=(4, 1024, 16, 64) (torch.float32)
    attn_bias : <class 'NoneType'>
    p : 0.0
    flshattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    requires a GPU with compute capability > 7.5
    tritonflashattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    triton is not available
    requires A100 GPU
    cutlassF is not supported because:
    xFormers wasn't build with CUDA support
    smallkF is not supported because:
    xFormers wasn't build with CUDA support
    max(query.shape[-1] != value.shape[-1]) > 32
    has custom scale
    unsupported embed per head: 64

显卡v100， 不太清楚是啥问题，可以在docker 镜像中更新可用的xformers吗

你好，之前docker镜像中确实没下xformers，因为作者配置文件中没写而且不需要xformers也能运行。而在该环境中安装xformers我也遇到了问题。目前我是在torch2.1.2+cu118的环境中配置好了xformers与ape的依赖，并上传了该镜像ape_cu118

我这没有xformer压根运行不了啊

@shutu-1
Copy link

shutu-1 commented Jun 6, 2024

keyk13/ape_cu118:v1

你之前没有xformer如何运行的呀，我这没有xformer压根运行不了呀

  1. 使用docker pull keyk13/ape_image:v1 拉取了在以上问题中提供的镜像
  2. 但是在容器中没有找到xformers库，pip install xformers 会安装0.0.23版本，自动更新torch版本；如果安装0.0.17版本，会有以下报错
    NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
    query : shape=(4, 1024, 16, 64) (torch.float32)
    key : shape=(4, 1024, 16, 64) (torch.float32)
    value : shape=(4, 1024, 16, 64) (torch.float32)
    attn_bias : <class 'NoneType'>
    p : 0.0
    flshattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    requires a GPU with compute capability > 7.5
    tritonflashattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    triton is not available
    requires A100 GPU
    cutlassF is not supported because:
    xFormers wasn't build with CUDA support
    smallkF is not supported because:
    xFormers wasn't build with CUDA support
    max(query.shape[-1] != value.shape[-1]) > 32
    has custom scale
    unsupported embed per head: 64

显卡v100， 不太清楚是啥问题，可以在docker 镜像中更新可用的xformers吗

你好，之前docker镜像中确实没下xformers，因为作者配置文件中没写而且不需要xformers也能运行。而在该环境中安装xformers我也遇到了问题。目前我是在torch2.1.2+cu118的环境中配置好了xformers与ape的依赖，并上传了该镜像ape_cu118

我这没有xformer压根运行不了啊

我使用的keyk13/ape_cu118:v1镜像，里面有
“xformers 0.0.23.post1+cu118 pypi_0 pypi”
环境呢，可能是你镜像要换成最新的，进“https://hub.docker.com/repository/docker/keyk13/ape_cu118”后点击tag，下载最新的v1镜像

@wdc233
Copy link

wdc233 commented Jun 7, 2024

keyk13/ape_cu118:v1

你之前没有xformer如何运行的呀，我这没有xformer压根运行不了呀

  1. 使用docker pull keyk13/ape_image:v1 拉取了在以上问题中提供的镜像
  2. 但是在容器中没有找到xformers库，pip install xformers 会安装0.0.23版本，自动更新torch版本；如果安装0.0.17版本，会有以下报错
    NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
    query : shape=(4, 1024, 16, 64) (torch.float32)
    key : shape=(4, 1024, 16, 64) (torch.float32)
    value : shape=(4, 1024, 16, 64) (torch.float32)
    attn_bias : <class 'NoneType'>
    p : 0.0
    flshattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    requires a GPU with compute capability > 7.5
    tritonflashattF is not supported because:
    xFormers wasn't build with CUDA support
    dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
    triton is not available
    requires A100 GPU
    cutlassF is not supported because:
    xFormers wasn't build with CUDA support
    smallkF is not supported because:
    xFormers wasn't build with CUDA support
    max(query.shape[-1] != value.shape[-1]) > 32
    has custom scale
    unsupported embed per head: 64

显卡v100， 不太清楚是啥问题，可以在docker 镜像中更新可用的xformers吗

你好，之前docker镜像中确实没下xformers，因为作者配置文件中没写而且不需要xformers也能运行。而在该环境中安装xformers我也遇到了问题。目前我是在torch2.1.2+cu118的环境中配置好了xformers与ape的依赖，并上传了该镜像ape_cu118

我这没有xformer压根运行不了啊

我使用的keyk13/ape_cu118:v1镜像，里面有 “xformers 0.0.23.post1+cu118 pypi_0 pypi” 环境呢，可能是你镜像要换成最新的，进“https://hub.docker.com/repository/docker/keyk13/ape_cu118”后点击tag，下载最新的v1镜像

image
链接不对

@shutu-1
Copy link

shutu-1 commented Jun 7, 2024

用这个
docker pull keyk13/ape_cu118:v1
直接拉

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
6 participants
@shenyunhang @Cristhine @kongdebug @zhiwenhou1227 @wdc233 @shutu-1