Error Data incompatible with tuples format. Each message should be a list of length 2.

[darkon12]

What should I do?

[unitycoder]

using Windows+Anaconda, same issue.
I added some print commands in the gradio method for debugging, its not receiving all data from the web form?

chatgipity™ made a workaround for me, at least it runs now and creates mesh:
modified chatbot.py (use some code compare/merge tool to see the modified parts and could remove the debug logs)
https://gist.github.com/unitycoder/ec21b35526bb78caac1fa002afc1d833

But of course, interested to know if there are better fixes..

[darkon12]

https://gist.github.com/unitycoder/ec21b35526bb78caac1fa002afc1d833

How do you use this?

[unitycoder]

replaced existing chatbot.py with that code (i think it was inside gradio package folder, might see the file location in console error messages)

[gazonk]

Thanks for the info. Something that seems to work is to set, in app.py, the chatbot's message type to "messages".
Here is the diff:

# Gradio block
-chatbot=gr.Chatbot(height=450, placeholder=PLACEHOLDER, label='Gradio ChatInterface')
+chatbot=gr.Chatbot(height=450, placeholder=PLACEHOLDER, label='Gradio ChatInterface', type="messages")

[gazonk]

I have done also modifications to avoid error related to padding token.
It would be exaggerated to say that I know exactly what I'm doing :-) but some spurious messages about inferring masking have disappeared :-). I'm interested in knowing what I'm doing wrong.
app.py.txt

[darkon12]

I have done also modifications to avoid error related to padding token. It would be exaggerated to say that I know exactly what I'm doing :-) but some spurious messages about inferring masking have disappeared :-). I'm interested in knowing what I'm doing wrong. app.py.txt

Uhm, no. The only thing that your mod does is to switch to a generic "error" message.
Your app.py.txt gets an OOM (al least in colab).
Will try again later.

[gazonk]

I probably don't get the OOM because I'm on a Macbook M2 with 96GB of ram.

[darkon12]

I probably don't get the OOM because I'm on a Macbook M2 with 96GB of ram.

Wow, bragging at its finest ;)

[gazonk]

I wouldn't brag about that ;-), it just explains why it works for me.