SBert model isn't selectable after download

[r-j-s]

Steps to Reproduce

1. Download SBert Model in "Discover and Download Models"
2. Close the dialog
3. Try to select the downloaded SBert Model, it seems like the list is clear

Your Environment

• Operating System: Windows 10 as well as Linux Mint 21.2
• Chat model used (if applicable): SBert

If I download other models everything work as expected, but the SBert model is not shown in the dropdown menu in GPT4ALL GUI. If I take look into the download folder there is a file called "all-MiniLM-L6-v2.gguf2.f16.gguf" successfully downloaded.

[karsten32]

I have the same problem: Downloaded SBert Model, but it wont show up in the dropdown menu so I cannot select it. Other models are working fine.

OP_System is windows 10 as well.

Any solutions or what did I do wrong?

BR

[cosmic-snow]

Let me explain: The SBERT model is not a chat model. It is used for the LocalDocs feature instead. Its description in the downloads list does mention that. And yes, its file name is all-MiniLM-L6-v2.gguf2.f16.gguf.

Did you expect to start a conversation after downloading it? There seems to be some confusion here. Can I ask what made you want to use it for chatting in the first place?

Also, looks you selected the bindings tag. Are you trying to program with the language bindings or is this only about the chat application?

[karsten32]

@cosmic-snow Thanks for quick reply!

Yes its confusing (for me). It appears that SBERT ist merely a kind of plugin for the LocalDocs feature and not a LLM Model which can be installed and used for interaction. However, SBERT is listed in the download section as "model" and thus I expected that I have to use the "model" SBERT when I want to interact/analyse my local files. The problem what causes confusion is that after one downloaded SBERT, SBERT does not appear anywhere in order to be activeted and/or selected for use. I assume that it is automatically activated when using the LocalDocs feature (?).

I tried it again and the LocalDocs feature works fine now.

@r-j-s

• Download the SBERT "module" / plugin (like you did) and further any LLM module which you want to use.

• go to "settings" (the gear icon on the top right) and there you go to "LocalDocs" and there you add the folder where you locally story the files you want to use. After that, exit the menu.

• then select the LLM module which you want to use for promting

• in order to make the chosen LLM module working with local files only, you have to go to the "three burger icon" on the top right (I don't know how to describe that symbol, but I am sure that you will find it). There you select the folder of your local files which shall be used. Then exit the menu and start prompting.

• while your prompted question is processed, there should appear "searching localdocs: [name of your folder selected]..." It should look like the following image ("LLM Private") is my local folder which I selected for use and in which I stored the documents which I wanted to use.

I watched this video on Youtube for more understanding which is, however, a bit outdated.

I hope this will help you.

BR

[cosmic-snow]

However, SBERT is listed in the download section as "model" and thus I expected that I have to use the "model" SBERT when I want to interact/analyse my local files. The problem what causes confusion is that after one downloaded SBERT, SBERT does not appear anywhere in order to be activeted and/or selected for use. I assume that it is automatically activated when using the LocalDocs feature (?).

Ah I see, that's where the confusion is. SBERT is a model, just not a Large Language Model (LLM) / Generative Pre-trained Transformer (GPT). Simplified, these two are what we call "chat models" (the former is a more general term and includes the latter).

SBERT is a kind of model which enables similarity search, which in turn is a requirement for the LocalDocs feature. So without SBERT or an alternative, LocalDocs in its current form won't work at all. SBERT does not produce any text whatsoever.

[karsten32]

@cosmic-snow

Thanks again for explanation.

BR