# Getting started — M.E.F LLM Studio

Source: https://www.mef-llm-studio.com/en/start/

Plan the setup before training. You can skip the training environment initially and read the learning chapters.

## What to prepare

- A Windows PC with an x64 processor.
- For GPU training, a suitable NVIDIA graphics card and compatible driver. The app checks eligibility during setup.
- Supported linked models can use a separate CPU inference profile; large models require correspondingly large system RAM.
- An internet connection for the optional private runtime environments; downloads may total several gigabytes.
- At least the free space shown in Setup. Initially allow a minimum of 10 GiB for the environment and caches.

## Read the alpha terms

The alpha is for non-commercial evaluation only. The installer requires acceptance of the EULA.

## Get to know the app

Open the learning chapters. The Visual Explorer requires Microsoft WebView2 Runtime. If it is missing, the app provides a Microsoft download link.

## Set up the training environment

Review packages, sources and disk space. Python and PyTorch download only after your separate confirmation.

## Try a linked model

Link a supported Safetensors model and choose CPU, NVIDIA GPU or A/B. Check its memory needs before loading.

## Start small

Use a small dataset and a short test run first. Batch size, sequence length and model size need to fit your graphics card.

## Back up and compare

Keep independent backups. Explore a checkpoint before planning larger runs.

Your work is not a temporary file.

The app restores inputs, chat history, form values and editor drafts after a restart. They are stored unencrypted in the local app-data area and can be cleared in Settings. Source files and A/B results are saved only when you choose to. The environment, models and datasets remain separate.

Contact: support@mef-llm-studio.com