Setting up the code and env

1. Prerequisites

1. Python 3.13.x must be installed - download
2. LM Studio v0.3.5 must be installed - release notes

2. Check that Python has been installed correctly

1. Execute following command

python --version

Output should look like below

Python 3.13.0

3. Setting up the LMStudio

In order to use the LMStudio as a server navigate to Developer Mode.

Set up the port to 8080 (or other - in case that 8080 is used)

4. Check that LMStudio can be run from commandline

lms
   __   __  ___  ______          ___        _______   ____
  / /  /  |/  / / __/ /___ _____/ (_)__    / ___/ /  /  _/
 / /__/ /|_/ / _\ \/ __/ // / _  / / _ \  / /__/ /___/ /
/____/_/  /_/ /___/\__/\_,_/\_,_/_/\___/  \___/____/___/

lms - LM Studio CLI - v0.0.28
GitHub: https://github.com/lmstudio-ai/lmstudio-cli

Usage
lms <subcommand>

where <subcommand> can be one of:

- status - Prints the status of LM Studio
- server - Commands for managing the local server
- ls - List all downloaded models
- ps - List all loaded models
- get - Searching and downloading a model from online.
- load - Load a model
- unload - Unload a model
- create - Create a new project with scaffolding
- log - Log operations. Currently only supports streaming logs from LM Studio via `lms log stream`
- import - Import a model file into LM Studio
- bootstrap - Bootstrap the CLI
- version - Prints the version of the CLI

For more help, try running `lms <subcommand> --help`

5. Clone the repository using git clone command:

git clone https://github.com/pwalaszkowski/msc_seminarium.git

$ git clone https://github.com/pwalaszkowski/msc_seminarium.git
Cloning into 'msc_seminarium'...
remote: Enumerating objects: 51, done.
remote: Counting objects: 100% (51/51), done.
remote: Compressing objects: 100% (35/35), done.
remote: Total 51 (delta 22), reused 41 (delta 12), pack-reused 0 (from 0)
Receiving objects: 100% (51/51), 41.93 KiB | 1.50 MiB/s, done.
Resolving deltas: 100% (22/22), done.

or download

https://github.com/pwalaszkowski/msc_seminarium/archive/refs/heads/master.zip

6. Navigate to repository directory

1. Setup Python venv in current working directory

python -m venv env

2. Confirm that env is created (the directory env should exist)

$ ls -la env/
total 14
drwxr-xr-x 1 pwala 197609   0 Dec 17 18:35 ./
drwxr-xr-x 1 pwala 197609   0 Dec 17 18:35 ../
-rw-r--r-- 1 pwala 197609  71 Dec 17 18:35 .gitignore
drwxr-xr-x 1 pwala 197609   0 Dec 17 18:35 Include/
drwxr-xr-x 1 pwala 197609   0 Dec 17 18:35 Lib/
drwxr-xr-x 1 pwala 197609   0 Dec 17 18:35 Scripts/
-rw-r--r-- 1 pwala 197609 337 Dec 17 18:35 pyvenv.cfg 

3. Navigate to directory env/Scripts

Activate virtual environment running the activate script

./activate.bat

Under Linux machine the script will be called

./activate 

4. The user prompt should now start with (env)

(env) C:\Repositories\

7. Download required packages via pip

Navigate to main directory (msc_seminarium)

Confirm that you are in main directory 

python -m pip install -r requirements.txt

Running prompts

1. Open LMStudio (do not run any prompts or download models) 2. Navigate to config.ini file to set the model, prompt and model paths under Windows

[DEFAULT]
MODEL = qwen2-0.5b-instruct
PROMPT = 'Create me a for loop in Python'
MODEL_PATH = C:\Users\pwala\.cache\lm-studio\models

3. Run the lmstudio_wrapper.py

   python ./lmstudio_wrapper.py

   The wrapper will
   1. launch the server
   2. Download model
   3. Load the model into LMStudio
   4. Send prompt to LMStudio
   5. Return answer on console
   6. Write the answer into output directory with following format MODEL_YYYY_MM_DD_HH_MM_SS
   7. The wrapper will eject the model
   8. The wrapper will remove the downloaded models

Cleanup the env. **

1. Navigate to directory env/Scripts. Deactivate virtual environment running the deactivate script

./deactivate.bat

Under Linux machine the script will be called

./deactivate 

2. The user prompt should now no longer start with (env)

C:\Repositories\

Model downloads

https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct

API

https://lmstudio.ai/docs/api/rest-api

Postman API Collection

Postman Collection