conda creating environment endlessly

[arch-user-france1]

Conda won't finish doing that:

I left it running for more than five hours and it still was "examining ..." something

I did:

1. To get a stable release
2. If this is not your first install, update Conda and remove old donkey
3. Create the Python anaconda environment (recommended faster install time)

https://docs.donkeycar.com/guide/host_pc/setup_ubuntu/

What could I do?

[arch-user-france1]

removing -n base -c conda-forge it worked, however now it says "mamba" command not found and mamba is not a python module either
I then tried pip install mamba but now I end up with
/usr/bin/python: No module named mamba.__main__; 'mamba' is a package and cannot be directly executed

[TCIII]

@arch-user-france1,

SBC/PC: ?
OS: ?

Was this your first install of conda, if so why did you perform #2?
I just installed mamba on my Xavier NX (JP502) running Ubuntu 20.04 without issue.

TCIII

[Ezward]

The ubuntu install documentation that you linked already has the answer I think;

• delete your previous environment

conda update -n base -c defaults conda
conda env remove -n donkey

• use mamba to install a new environment (from within the donkeycar project folder)

conda install mamba -n base -c conda-forge
mamba env create -f install/envs/ubuntu.yml
conda activate donkey
pip install -e .[pc]

[arch-user-france1]

That's what I did but it won't install mamba that way, just stuck

[TCIII]

@Ezward,

I believe that arch-user-france1's command line should say (base) when installing mamba, if conda has been installed correctly, and it doesn't so has conda been installed correctly?

TCIII

[arch-user-france1]

@Ezward,

I believe that arch-user-france1's command line should say (base) when installing mamba, if conda has been installed correctly, and it doesn't so has conda been installed correctly?

TCIII

Conda has been installed correctly. There is no wrong way.

[sctse999]

I support this finding. Conda takes a lot of time to resolve dependencies and sometimes it is taking forever. I am suspecting that this is due to pytorch or other depdencies.

@arch-user-france1 Could you please comment out the following and get back?

In ubuntu.yml (https://github.com/autorope/donkeycar/blob/main/install/envs/ubuntu.yml)

delete the following lines:

• pytorch=1.7.1
• torchvision
• torchaudio
• pytorch-lightning
• fastai

Delete the donkey environment and retry the installation.

[TCIII]

@sctse999,

I just recently installed Conda on a Xavier NX 8GB running JP502 and I selected the Python 3.8 version of conda, for ARM processors, which installed without issues.
When I went to install mamba, the CLI indicated (base), which is the correct env prior to creating the (donkey) env, unlike arch-user-france's CLI which displays "donkey",
Mamba installed without issue though we had to adjust the ubuntu.yml and [pc] for a successful install of DC.

TCIII

[arch-user-france1]

I support this finding. Conda takes a lot of time to resolve dependencies and sometimes it is taking forever. I am suspecting that this is due to pytorch or other depdencies.

@arch-user-france1 Could you please comment out the following and get back?

In ubuntu.yml (https://github.com/autorope/donkeycar/blob/main/install/envs/ubuntu.yml)

delete the following lines:

* pytorch=1.7.1

* torchvision

* torchaudio

* pytorch-lightning

* fastai

Delete the donkey environment and retry the installation.

the problem is it's hanging here: conda install mamba -n base -c conda-forge

[sctse999]

@arch-user-france1 try not to use mamba and skip this line. In the next line, use this:

conda env create -f install/envs/ubuntu.yml

[sctse999]

@sctse999,

I just recently installed Conda on a Xavier NX 8GB running JP502 and I selected the Python 3.8 version of conda, for ARM processors, which installed without issues. When I went to install mamba, the CLI indicated (base), which is the correct env prior to creating the (donkey) env, unlike arch-user-france's CLI which displays "donkey", Mamba installed without issue though we had to adjust the ubuntu.yml and [pc] for a successful install of DC.

TCIII

Unfortunate it doesn't work for me today on both Windows 11 and MacOS (Apple M1/M2). All of them are stuck in somewhere and cannot proceed.

[TCIII]

@sctse999,

Interesting to say the least.

It took a lot of flag waving to get the ubuntu.yml and [pc} to run. I suggest that you get on the DC Discord channel and send DGarbanzo a private message explaining your issues and I believe he can help you.

TCIII

[arch-user-france1]

@arch-user-france1 try not to use mamba and skip this line. In the next line, use this:

conda env create -f install/envs/ubuntu.yml

so then we're doing

Alternative (slower install time):

which would be basically never ending

[arch-user-france1]

I support this finding. Conda takes a lot of time to resolve dependencies and sometimes it is taking forever. I am suspecting that this is due to pytorch or other depdencies.

@arch-user-france1 Could you please comment out the following and get back?

In ubuntu.yml (https://github.com/autorope/donkeycar/blob/main/install/envs/ubuntu.yml)

delete the following lines:

* pytorch=1.7.1

* torchvision

* torchaudio

* pytorch-lightning

* fastai

Delete the donkey environment and retry the installation.

trying this one right now; still solving environment, but we'll see if it's faster. Will respond once either complete or "hanging".

[arch-user-france1]

still solving environment, so fastai might not be the problem

[TCIII]

@sctse999,

When we installed ubuntu.yml we actually used jetson.yml, which is here, which removes the pytorch dependencies and modifies kivy.

TCIII

[sctse999]

still solving environment, so fastai might not be the problem

Have you tried deleting pytorch and its friends too?

[arch-user-france1]

I did what you have asked for and nothing more

[sctse999]

@arch-user-france1 so you have removed pytorch as well?

[sctse999]

Can you use main instead of stable release tag?

[arch-user-france1]

Can you use main instead of stable release tag?

I didn't use stable last time

[ssh-keyz]

I'm having this same issue on an M1 Mac on the master branch

[ssh-keyz]

I was able to get the environment running.
I removed the packages mentioned above by @sctse999
I ran this command conda config --set restore_free_channel true
Created the env using Rosetta: CONDA_SUBDIR=osx-64 conda create -f install/envs/mac.yml

[sctse999]

I was able to get the environment running. I removed the packages mentioned above by @sctse999 I ran this command conda config --set restore_free_channel true Created the env using Rosetta: CONDA_SUBDIR=osx-64 conda create -f install/envs/mac.yml

@tylerjheslop Thanks for the information.

what does conda config --set restore_free_channel true do? What's its effect?

I am not using Mac. Is it a must to use Rosetta?