collections is not a package about fastsurfer HOT 12 CLOSED

In addition, the https://github.com/Deep-MI/FastSurfer/blob/master/Docker/Dockerfile has many python packages in it. Are they all present in requirements.txt? For example, I don't see cudatoolkit=10.0 "pytorch=1.2.0=py3.6_cuda10.0.130_cudnn7.6.2_0" torchvision=0.4.0 in requirements.txt.

from fastsurfer.

Potentially fixed by PR #18

from fastsurfer.

Even bigger trouble, why does requirements.txt environment don't use GPU at all? Is it because of those missing packages?
@LeHenschel
In the former environment, I tried Example 1 and that is going on for 30 minutes now. top shows 800% CPU usage while nvidia-smi does not show any GPU usage, which in turn supports my finding.

from fastsurfer.

It looks like you made requirements.txt from https://github.com/Deep-MI/FastSurfer/blob/master/Docker/Dockerfile_CPU#L67, not the GPU one.

from fastsurfer.

i meet the same question, Could not find a version that satisfies the requirement collections, could you tell me how to solve this question? thank you @tashrifbillah

from fastsurfer.

@ClePol @m-reuter

from fastsurfer.

Sure, please build the requiresments.txt from my PR #18.

from fastsurfer.

To be more specific, see my branch and follow the installation instruction I edited.
https://github.com/tashrifbillah/FastSurfer/tree/fix-requirements

from fastsurfer.

To be more specific, see my branch and follow the installation instruction I edited.
https://github.com/tashrifbillah/FastSurfer/tree/fix-requirements

Wow, I ran the program successfully using the method you provided. Thank you very much

from fastsurfer.

You are welcome @congzhaoyang , for whatever reason my PR didn't get any attention yet :)

from fastsurfer.

Hi, we are aware of this and will fix this in September (due to vacation time). Thanks

from fastsurfer.

Hey,

thank you for bringing this to our attention. I removed the unnecessary default packages (collections etc.) and included an extended description regarding the different pytorch version you can install.

As a matter of fact, the requirements.txt file does represent the GPU version of FastSurfer as can be seen here: https://github.com/Deep-MI/FastSurfer/blob/master/requirements.txt#L24

By default, pip will install torch v1.2.0 gpu version for CUDA 10.0. For the CPU version, this would be changed by adding +cpu to the end of the version definition for both torch and torchvision. Similarly, if you need a different CUDA version (e.g. if your NVIDIA-Driver only supports an older version) you have to specify this via the tag as well (see also instructions for v1.2.0 (Wheel) at: https://pytorch.org/get-started/previous-versions/#wheel-4):

# CUDA 10.0
pip install torch==1.2.0 torchvision==0.4.0

# CUDA 9.2
pip install torch==1.2.0+cu92 torchvision==0.4.0+cu92 -f https://download.pytorch.org/whl/torch_stable.html

# CPU only
pip install torch==1.2.0+cpu torchvision==0.4.0+cpu -f https://download.pytorch.org/whl/torch_stable.html

Now, for conda this process is done differently. Here, you specify the desired version via the cudatoolkit option, which is what we do in the Docker (see instructions for v1.2.0 (Conda) https://pytorch.org/get-started/previous-versions/#conda-4):

# CUDA 9.2
conda install pytorch==1.2.0 torchvision==0.4.0 cudatoolkit=9.2 -c pytorch

# CUDA 10.0
conda install pytorch==1.2.0 torchvision==0.4.0 cudatoolkit=10.0 -c pytorch

# CPU Only
conda install pytorch==1.2.0 torchvision==0.4.0 cpuonly -c pytorch

Hence, cudatoolkit=10.0 is not needed in requirements.txt. I now also added two yml-files for the gpu and cpu version of fastsurfer in which you can see the cudatoolkit defintion.

I hope this clears thing up and helps you running things smoothly.

Best,
Leonie

from fastsurfer.