Summary of how to use lightweight chat AI 'GPT4ALL' that can be used even on low-spec PCs without Grabo

High-performance chat AIs, such as ChatGPT, are being announced one after another, but most chat AIs need to be used online, and most chat AIs that operate offline require high-performance PCs. Under such circumstances, '

GPT4ALL ' has appeared, which can be operated even on a PC that does not have a high-performance GPU. When I actually used it, I was able to run it on a mobile notebook PC that does not have a graphic board, so I tried to summarize how to use it.

GitHub - nomic-ai/gpt4all: gpt4all: a chatbot trained on a massive collection of clean assistant data including code, stories and dialogue

GPT4ALL is a chat AI developed by Nomic AI , featuring high-precision conversational ability and light enough to run on a laptop. The following article explains in detail what kind of AI GPT4ALL is. This time, I will check the procedure to actually use GPT4ALL on Windows.

Announcement of 7 billion parameter chatbot 'GPT4ALL' that can be run on laptops for free - GIGAZINE

As mentioned above, GPT4ALL features a light weight that can run on a notebook PC. So, this time I will try using GPT4ALL with a mobile notebook PC ' VAIO SX12 ' that does not have a graphic board.

First, access the link below to download the necessary files.

GitHub - nomic-ai/gpt4all: gpt4all: a chatbot trained on a massive collection of clean assistant data including code, stories and dialogue

Next, click 'Code' and then 'Download ZIP' to download the ZIP file (file size: 3.54MB).

After downloading the ZIP file, decompress it using Windows standard functions or software such as Explzh .

Once the file 'gpt4all-main' is created, double-click it to open it.

It is OK if the folders and files are arranged in a row as shown below.

Next, download the GPT4ALL model data (file size: 3.92GB) by clicking the direct link below or using a torrent magnet link. In order to avoid server congestion, it is better to use magnet links if you have an environment that can handle torrents.

Direct link:

Magnet Link (Torrent software required): magnet:?xt=urn:btih:1F11A9691EE06C18F0040E359361DCA0479BCB5A&dn=gpt4all-lora-quantized.bin&

After downloading the model data, move it into the 'chat' folder inside the 'gpt4all-main' folder.

It is OK if the model data is placed in the 'chat' folder like this.

Next, display the top of the 'gpt4all-main' folder (one level above the 'chat' folder) in Explorer and click the blank area in the address bar at the top of the screen.

When you can enter text, enter 'powershell' and press Enter.

Then Powershell will start with the 'gpt4all-main' folder open.

Once Powershell starts, run the following commands:
[code]cd chat;./gpt4all-lora-quantized-win64.exe[/code]

An image showing how to execute the command looks like this. First, just copy and paste the entire command above and press Enter.

After executing the command, the chat AI will start. When '>' is displayed at the bottom of the screen, you can start a conversation.

The conversation method is as follows. First, type a sentence and press Enter.

After waiting for a while, the AI will respond.

Below is the screen when asking how to make curry and drinks that match curry. It will give you the right answer for your question.

The result of entering Japanese is as follows. Even if I enter Japanese, a square symbol is displayed.

As a result of using 'Windows Terminal' instead of Powershell, Japanese input was successful, but the response 'Cannot complete the process' was returned in English.

As a result of using GPT4ALL on macOS, it responded in Japanese to Japanese input. However, the content of the reply is incoherent. GPT4ALL is fully capable of communicating in English, but it is quite difficult to support Japanese.

The CPU usage rate while AI is executing processing is like this. All threads are stuck at around 100%, and you can see that the CPU is being used to the maximum. On the other hand, if you focus on the GPU usage rate on the left side of the screen, you can see that the GPU is hardly used. Since GPT4ALL does not require GPU power for operation, it can be operated even on machines such as notebook PCs that do not have a dedicated graphic board.

in Review,   Software, Posted by log1o_hf