[D-96] Mamba Environment

Mamaba Environment

[1] Create a new environment:

conda create -n mamba python=3.10

[2] activate:

conda activate mamba

[3] install pytorch:

conda install pytorch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 pytorch-cuda=11.8 -c pytorch -c nvidia

[4] load cuda module:

module load cuda/11.8

[5] install causal-conv1d: error

pip install causal-conv1d

Try 1

pip install git+https://github.com/Dao-AILab/causal-conv1d

Try 2

conda install conda-forge::causal-conv1d

::» worked!

$ conda list causal-conv1d
# packages in environment at /storage/athene/work/kim/miniconda3/envs/pytorch310:
#
# Name                     Version          Build                   Channel
causal-conv1d              1.5.4            cuda129py310ha1a813f_0  conda-forge

[6] install mamba-ssm: error

pip install mamba-ssm==2.2.2

Try 1

conda install conda-forge::mamba-ssm

::» worked!

$ conda list mamba-ssm
# packages in environment at /storage/athene/work/kim/miniconda3/envs/pytorch310:
#
# Name                     Version          Build            Channel
mamba-ssm                  2.2.5            py310he9a7d99_1  conda-forge

I hope conda-forge channel doesn’t cause any problem further.

TODO Tomorrow

causal-conv1d and mamba-ssm are installed under pytorch310 conda environment

1. Explore Shared folder.
2. Get involved with pre-training code and its dataset.