fairseq audio model

#6
by tanaythatte - opened

I have a fairseq audio model in .pt format how can i train ultravox using that audio model

Fixie.ai org

@tanaythatte please take a look at this script we use for creating the continuation training datasets: https://github.com/fixie-ai/ultravox/blob/main/ultravox/tools/ds_tool/ds_tool.py

Fixie.ai org

Sorry that seems like the wrong response.

I don't know if there's a general solution to that question. Depending on the model and the differences that can be very easy to very hard. You probably need to convert it to a format that HF understands or manually change the code. There might also be special converters for example someone provided this one here: https://huggingface.co/fixie-ai/ultravox-v0_4/discussions/5#676fb4cc8756741d47dadf59

Fixie.ai org

@farzadab thanks for catching this. Somehow I misread the question.

Thanks for replying
I tried the script but getting some errors
Do you have any more sources

Sign up or log in to comment