Cannot create Quiz with Olama

[Sauron-sol]

Hello,

I'm experiencing an issue when trying to create a quiz using Olama. After selecting my file, I encountered the following error: "Cannot read properties of undefined (reading 'forEach')."
My file is located in a folder, so I also tried using both the "Add note" and "Add folder" options to test if there was a bug, but the issue persists.

Could you please provide guidance on how I can successfully create a quiz?
Thanks

[ECuiDev]

Do you get the error when you try to add a note/folder or when you try to generate questions?

[Sauron-sol]

When I try to generate questions

[ECuiDev]

Do you have Ollama installed and running when you try to generate questions? If so, what generation model are you using?

[LeonelRFF]

Hi, do you have ollama pull nomic-embed-text installed? It is necessary for it to work, just in case if you don't have it you should open cmd or powershell and install it

ollama pull nomic-embed-text

Since I'm passing by it seems that just in this case a bug occurred 😁 as you can see in the video the options card to join cannot be displayed I realized that it is because for this case as you can see in the callout it is like this

quiz.mp4

> [!question] El contenido estructurado se disfruta más porque cada párrafo tiene que estar delimitado por un elemento de tipo _______________
> a) <b>
> b) <i>
> c) <p>
> d) <h1>
>> [!success]- Answer
>> c) <p>

if I change <h1> or the rest of the options for another word it works correctly

[ECuiDev]

Hi, do you have ollama pull nomic-embed-text installed? It is necessary for it to work, just in case if you don't have it you should open cmd or powershell and install it

ollama pull nomic-embed-text

Since I'm passing by it seems that just in this case a bug occurred 😁 as you can see in the video the options card to join cannot be displayed I realized that it is because for this case as you can see in the callout it is like this

quiz.mp4

> [!question] El contenido estructurado se disfruta más porque cada párrafo tiene que estar delimitado por un elemento de tipo _______________
> a) <b>
> b) <i>
> c) <p>
> d) <h1>
>> [!success]- Answer
>> c) <p>

if I change <h1> or the rest of the options for another word it works correctly

This is because the actual HTML is being rendered. I'll add a generation-level fix for this in the future, but for now you should enclose HTML with backticks so that it gets treated as text.

> [!question] El contenido estructurado se disfruta más porque cada párrafo tiene que estar delimitado por un elemento de tipo `_______________`.
> a) `<b>`
> b) `<i>`
> c) `<p>`
> d) `<h1>`
>> [!success]- Answer
>> c) `<p>`

[Sauron-sol]

Yes, Ollama is running and here are the models I'm running for generation and embedding (when I type "ollama ps", my two models are running) I also tried with llama3.2:latest, but it doesn't work (of course I have "nomic-embed-text:latest" installed)

[ECuiDev]

Hmm, that's very strange. Is it possible for you to send a video of what happens when you try to generate a quiz?

[Sauron-sol]

QuizGenerator.mp4

Here's the video

[ECuiDev]

What happens if you try generating using a small note (few hundred tokens)?

[Sauron-sol]

It seems to work, but not over 1000 tokens. I tried it with 349 tokens and it worked

[ECuiDev]

I believe the hardware affects the model speed, so maybe if your machine isn't powerful enough and you make a really large request, the model could timeout? That feels like the most likely scenario to me right now.

You could also try asking in the Ollama discord server if anyone has encountered a similar problem when sending a large number of tokens in a single request.

[SeamusMullan]

I've been having this same issue with the following specs + settings

Apple Macbook Air M2 (2022) 8GB Ram
MacOS Sequoia 15.0.1

I've noticed this issue also only seems to present itself when theres more that 1000 tokens making me think its an Ollama limitation. Perhaps splitting multiple pages into chunks and processing them seperately could be a temporary fix for this if there's no solution in the Ollama discord that's known of

[ECuiDev]

I see. I'll look into a fix.

[NotPubi]

Hello, i am having this same issue, any ideas?

[ECuiDev]

You get the same error when using more than 1000 tokens?

[NotPubi]

It happens even with 300 tokens. I haven't been able to generate a quiz yet. I tried a few different notes and models and I got the same thing

…

On Tue, Oct 15, 2024, 11:58 PM Edward Cui ***@***.***> wrote: You get the same error when using more than 1000 tokens? — Reply to this email directly, view it on GitHub <#19 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APLHISWKNKJDUKXQ2FJEVALZ3XW53AVCNFSM6AAAAABPQCNG6KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMJVG4ZTEMRYGU> . You are receiving this because you commented.Message ID: ***@***.***>

[ECuiDev]

How many questions did you try generating? And what models did you try?

[NotPubi]

I tried to generate 15 questions using llama 3.2:3b and 3.1, phi 3.5. I tried both nomic text embed and mxbai. Is there some way to debug it and maybe I can give you better information?

…

On Wed, Oct 16, 2024, 6:41 PM Edward Cui ***@***.***> wrote: How many questions did you try generating? And what models did you try? — Reply to this email directly, view it on GitHub <#19 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APLHISSO33GYDMPA5NHV6SDZ332THAVCNFSM6AAAAABPQCNG6KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMJYGE2DMNZTHA> . You are receiving this because you commented.Message ID: ***@***.***>

[ECuiDev]

I would try starting from a working threshold and go from there to see what breaks it. For example, generating 1 true or false question using 200 tokens. Then increasing either the number/type of questions or number of tokens.

[SeamusMullan]

Is there a way we could split the input info into chunks and generate questions with multiple calls to Ollama?

Something like every 800-1000 tokens since that seems to be the limit people are hitting before they get an error