Review of 'LM Studio', an execution environment that allows you to test the performance of various language models without the need for commands even on a non-GPU-equipped laptop

Research and development of text generation AI is progressing rapidly, and many large-scale language models (LLMs) that can be used for free are now available. However, many people give up from the start because it often requires high-performance hardware and complex environment construction to actually try out the operation of LLMs. By using the free ' LM Studio ', it is said that you can easily run LLMs without commands even on a PC with average performance, so I actually used LM Studio to check the steps to use LLM.
LM Studio - Discover, download, and run local LLMs
◆Installation of LM Studio
LM Studio is available for Windows and macOS, and there is also a beta version for Linux. The only requirement for using it on Windows is a CPU that supports AVX2. Most PCs sold in recent years are equipped with an AVX2-compatible CPU, so it can run in a wide range of environments. To install the Windows version of LM Studio, first access the link below and download the installer.
LM Studio - Discover, download, and run local LLMs
https://lmstudio.ai/
Click 'Download LM Studio for Windows' to download the installer (file size 401MB).

Double-click the installer to launch it.

If you see the following warning, click Run.

The installation process will proceed automatically without you having to click anything, and LM Studio will start automatically after the installation is complete. When you first start LM Studio, a summary screen will appear, so click 'Close' to close it.

You are now ready to use LM Studio.
◆How to run LLM in LM Studio
LM Studio allows you to download, manage, and run LLMs in one app. This time, try downloading and running the Japanese-compatible LLM '

The search results displayed the derivative LLMs of 'ELYZA-japanese-Llama-2-7b-instruct' available in LM Studio. After selecting the LLM, click 'Download' to download the LLM. In this case, click '

Wait for a while until the download is complete. The download progress will be displayed at the bottom of the screen.

Once the LLM download is complete, click the speech bubble button on the left side of the screen to switch to the chat screen.

When the chat screen appears, click “Select a model to load” at the top of the screen to specify the LLM to use in the chat, and then click “ELYZA japanese Llama 2 instruct” in the list of LLMs.

The status screen at the bottom of the screen is distracting, so click the X button to hide it.

When the chat input field hidden in the status screen appears, enter any text and press Enter. In this example, I entered 'What is the capital of Japan?'

After waiting for a while, the answer came back: 'Tokyo is the capital of Japan.'

We were also able to quickly run tests such as 'Can it compose haiku?' and 'Can it suggest recipes?'

When we checked the system load while the AI was generating text, we found that the CPU was running at full capacity, but the GPU was barely using it.


Related Posts: