search

huggingface / transformers

🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
https://huggingface.co/transformers
Apache License 2.0
134.79k stars 26.96k forks source link

Architectures for Dialogue #2562

Closed lukasfrank closed 4 years ago

lukasfrank commented 4 years ago

❓ Questions & Help

Hi 👋 I'm trying to build a dialogue system which should reply based on a history, memory (which is represented as a string) and a confidence if the memory content is correct and should be used.

Here two examples:

First of all, are there some best practices how to encode non textual input or inject such information into a model?

I already trained a Bert2Bert model which is not really performing very well. The generated response seems to be not conditioned on the encoder output.

Are there any recommendations what to try next?

Many thanks in advance for your hints!

stale[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

julien-c commented 4 years ago

Did you check out DialoGPT by @dreasysnail?

stale[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.