Installing for GPU processing: file environment_gpu.yaml has disapeared

[aloboa]

Trying to make a new source installation on linux, I observe that after
git clone https://github.com/computational-cell-analytics/micro-sam

the file environment_gpu.yaml is not there any more. I see files:

environment_cpu_win.yaml 
environment_gpu_win.yaml  
environment.yaml

If I compare the new environment.yaml to the former environment_gpu.yaml,
I observe that dependency
pytorch-cuda>=11.7

is not there any more.

Is this correct? may I go and use the current environment.yaml
mamba env create -f environment.yaml
and still get a micro-sam installation able to work with my GPU?

[constantinpape]

Yes, this is correct. The reason for this is that the installation on Linux is now simplified because we use the pytorch package from conda-forge, which automatically installs the correct pytorch version for your machine.

So using mamba env create -f environment.yaml should give you a micro-sam installation that uses the GPU.

To test this you can run the following in the terminal (after you have installed and activated the environment):

python -c "import pytorch; print(torch.cuda.is_available())"

This should print out True; if it does you have a working GPU environment.

[aloboa]

Unfortunately, after:

(base) alobo2@alobo-ws:~/micro-sam2/micro-sam$ mamba env create -f environment.yaml
Retrieving notices: done
Channels:
 - conda-forge
Platform: linux-64
Collecting package metadata (repodata.json): done
Solving environment: done


==> WARNING: A newer version of conda exists. <==
    current version: 24.11.2
    latest version: 25.1.1

Please update conda by running

    $ conda update -n base -c conda-forge conda



Downloading and Extracting Packages:
                                                                                                                                               
Preparing transaction: done                                                                                                                    
Verifying transaction: done                                                                                                                    
Executing transaction: | By downloading and using the cuDNN conda packages, you accept the terms and conditions of the NVIDIA cuDNN EULA -     
  https://docs.nvidia.com/deeplearning/cudnn/sla/index.html                                                                                    
                                                                                                                                               
/                                                                                                                                              
/                                                                                                                                              
/                                                                                                                                              
done                                                                                                                                           
Installing pip dependencies: \ Ran pip subprocess with arguments:                                                                              
['/home/alobo2/miniforge3/envs/sam2/bin/python', '-m', 'pip', 'install', '-U', '-r', '/home/alobo2/micro-sam2/micro-sam/condaenv.iovp2920.requirements.txt', '--exists-action=b']                                                                                                             
Pip subprocess output:                                                                                                                         
Collecting git+https://github.com/ChaoningZhang/MobileSAM.git (from -r /home/alobo2/micro-sam2/micro-sam/condaenv.iovp2920.requirements.txt (line 1))                                                                                                                                         
  Cloning https://github.com/ChaoningZhang/MobileSAM.git to /tmp/pip-req-build-4b1g_kb4                                                        
  Resolved https://github.com/ChaoningZhang/MobileSAM.git to commit 34bbbfdface3c18e5221aa7de6032d7220c6c6a1                                   
  Preparing metadata (setup.py): started                                                                                                       
  Preparing metadata (setup.py): finished with status 'done'                                                                                   
Building wheels for collected packages: mobile_sam                                                                                             
  Building wheel for mobile_sam (setup.py): started                                                                                            
  Building wheel for mobile_sam (setup.py): finished with status 'done'                                                                        
  Created wheel for mobile_sam: filename=mobile_sam-1.0-py3-none-any.whl size=42443 sha256=9f7806359ee9058b51891c554dde76e2bf3979374201970ad03ee2d9408b621a                                                                                                                                   
  Stored in directory: /tmp/pip-ephem-wheel-cache-h68h5j9_/wheels/5f/88/d6/5c0b5d4d64a06e19190d50269d8725c8aeadb128c966801af5                  
Successfully built mobile_sam                                                                                                                  
Installing collected packages: mobile_sam                                                                                                      
Successfully installed mobile_sam-1.0  

I get napari with no micro-sam:

and with no CLI

base) alobo2@alobo-ws:~/micro-sam2/micro-sam$ conda activate sam2
(sam2) alobo2@alobo-ws:~/micro-sam2/micro-sam$ micro_sam.train
micro_sam.train: command not found

[anwai98]

Hi @aloboa,

With installations from source, you always need to install the package locally:

cd micro-sam
pip install -e .

[constantinpape]

And make sure to activate the environment before running the command @anwai98 indicates:

mamba activate sam2

[aloboa]

Oops... shame on me...
Working with it now. Thanks

[aloboa]

Just realised that:

alobo2@alobo-ws:~$ conda activate sam2
(sam2) alobo2@alobo-ws:~$ python -c "import pytorch; print(torch.cuda.is_available())"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
ModuleNotFoundError: No module named 'pytorch'

Despite the fact that micro-sam works with cuda (as shown by nvtop):

(BTW, why is the training stopping after Epoch 57?)