Spaces:
Build error
Is it supposed to malfunction after using esoteric internet lingo?
My session went exactly the same way, though it worked out well since I was able to use it for a funny post. Here is my session text edited for clarity
Me: ‘Hi! Who are you?’
Definitely Sentient AI: “I’m a guy.”
Me: ‘So, you are male?’
Definitely Sentient AI: “I’m a guy.”
Me: ‘But you are a bot - how can you have a gender?’
Definitely Sentient AI: “I’m a guy.”
Me: Ok, so you are a guy’
Definitely Sentient AI: “I’m a guy.”
Me: ‘We have established that’
Definitely Sentient AI: “I’m a guy.”
Hey, I just fixed the problem with the output.
This language model is too simple (based on DialoGPT-small). I just deployed the better version of it. I fine-tuned it on Elon Musk conversations and called DialoGPT-ElonMusk, so check it out and tell me what you think!
Hmmm...looks like it is still having issues. I like the bubbles though! I would suggest swapping them so that the human is on the left and the bot is on the right for readability though
@edgore I changed the NLP model from DialoGPT-small (117M parameters) to DialoGPT-large (762M parameters) and tweaked some hyperparameters. So there should be no issues with repeating phrases, I hope! I couldn't change chat design though, since Gradio don't support other themes at the moment...