sciserver install ollama

sciserver_install_ollama.sh

#!/bin/bash

# Step 1: Create a directory called "ollama"
mkdir -p ollama

# Step 2: Inside the "ollama" directory, create a directory called "models"
mkdir -p ollama/models

# Step 3: Download the file into the "ollama" directory
cd ollama
curl -L https://ollama.com/download/ollama-linux-amd64.tgz -o ollama-linux-amd64.tgz

# Step 4: Install ollama by extracting the tar file in the current directory
tar -C ./ -xzf ollama-linux-amd64.tgz

# Step 5: Download the 'update_ollama_vars.sh' script directly
curl -L https://gist.githubusercontent.com/imaitland/4408f82852c41484712c287b778f07f8/raw/da9c040bc55e88b34bd677f5061d5e709440b72b/sciserver_update_ollama_vars.sh -o update_ollama_vars.sh

# Step 6: Source the 'update_ollama_vars.sh' script
source update_ollama_vars.sh

# End of script

some notes about sciserver's file system.

• ~/workspace/Storage/<username>/persistent has 10gb capacity and is backed up to tape!, not really the right spot for this.

• ~/workspace/Temporary/imaitla1/scratch/jobs is unlimited capacity but limited persistence, ~90days. Not a huge issue as installing ollama once you've figured it out isn't too bad.

• To consider, I (iain) don't have write permissions on the /ssec dir you can choose to mount when you create your container. Once I have write permissions there I'll look at creating a dedicated ollama dir, with associated bin and models there.

instructions
in sciserver's jupyter lab ui, create a file install_ollama.sh at the path described above, copy this gist's contents there. Then in the jupyter lab terminal run bash install_ollama.sh.

note: this install script copies the script located here to update env vars so you can run ollama's installed binary from the CLI, even though it's at an exotic path.