search

kipoi / kipoi-veff

Variant effect prediction plugin for Kipoi
https://kipoi.org/veff-docs
MIT License
6 stars 5 forks source link

compatability issues with installation of deepsea #44

Open cmlakhan opened 3 years ago

cmlakhan commented 3 years ago

Screen Shot 2021-01-25 at 5 43 29 PM I am trying to install Kipoi-Veff in conjunction with DeepSea and seem to be having problems. I am using an AWS instance with the deep learning AMI. I simply activate the python3 + cuda 10.0 environment and then install kipoi-veff using conda. This seems to install kipoi and kipoi-veff but then when I then try to install DeepSea into the same environment

kipoi env install DeepSEA/predict --gpu

I am getting compatibility issues which eventually break things.

If I am trying to install everything from scratch, how would you recommend I install Kipoi, Kipoi-veff, Basenji, and DeepSea?

haimasree commented 3 years ago

Hi Chirag, Short answer: Sorry no. I tried installing kipoi-veff with pip install kipoi-veff inside the docker container haimasree/kipoi-docker:sharedpy3keras2 (cpu version) and it seems to install fine. I am attaching the conda environment if it is of any help. This docker image is based on continuumio/miniconda3:4.8.2 which in turn I think is based on debian. So, if your flavor of os is drastically different then these libraries in the environment file will most likely not be available. Maybe, some of the pinned package versions help?

Further, to elaborate, I have been working on making kipoi models more robust for about 3/4 months. During which I created the docker images containing cpu only conda environments covering the 29 model groups in Kipoi. More information is available here - https://github.com/haimasree/kipoi-containers. I have plans to containerize the gpu version too shortly. I only started looking into kipoi-veff like this week, We are planning to improve the usability of kipoi-veff to avoid exactly the kind of problem you are facing right now. Hopefully it will get resolved soon but I am afraid we are just in the initial planning phase :)

On another note - Clusters which do not allow docker but allows singularity can use singularity to convert the docker containers to singularity containers. See the "singularity support" section in and https://github.com/haimasree/kipoi-containers and https://github.com/haimasree/kipoi-containers/blob/main/build-singularity-container.sh#L24-L25

Thank you for reporting the issue. I am sorry I dont have a solution just yet but I will be sure to post it if I have one.

Regards, Haimasree

On Fri, Feb 5, 2021 at 2:47 PM Chirag Lakhani notifications@github.com wrote:

Yeah that could be helpful. I am using an HPC cluster at my institution so I will have to see if they support docker. Are you saying you have DeepSea, Basenji, and Kipoi-Veff installed on a single docker instance? I have problems with Kipoi-veff interacts with the main kipoi installation.

Chirag

On Thu, Feb 4, 2021 at 10:19 AM Haimasree notifications@github.com wrote:

Will you be interested in exploring docker containers for this purpose? I have the cpu versions of Basenji and DeepSea tested with this docker image

docker pull haimasree/kipoi-docker:sharedpy3keras2

If you would like, I can create gpu friendly docker images as well. If you need kipoi separately,

docker pull haimasree/kipoi-docker:kipoi-base-env

I have not worked with kipoi-veff yet but as discussed I have started looking into it.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kipoi/kipoi-veff/issues/44#issuecomment-773386605, or unsubscribe < https://github.com/notifications/unsubscribe-auth/AAQMA5W54LUZRRTA6QRQO73S5K3HHANCNFSM4WSNY6WQ

.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kipoi/kipoi-veff/issues/44#issuecomment-774042775, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHMRTVAMUPT7BA7TFRKLX4DS5PZHTANCNFSM4WSNY6WQ .