Heewoo Jun

It's hard to say without looking at the exact commands, but these usually mean that some flags used during training are missing at sampling time

The easiest approach would be [to train your own top-level prior on a new dataset](https://github.com/openai/jukebox#reuse-pre-trained-vq-vae-and-retrain-top-level-prior-on-new-dataset). In theory, if you have enough VRAM/GPUs, you could finetune from our pretrained top-level priors,...

^ 👍 also, [this function](https://github.com/openai/jukebox/blob/086f90b07095db28fe4436f876aef50053aac499/jukebox/data/files_dataset.py#L72-L75) has to be implemented if you'd like to enable `--labels=True`. But, if there's only one that you'd like to finetune from, you can actually just...

We've updated the instructions on how to finetune from 1b_lyrics or train from scratch. Hope it helps!

Yeah, we didn't support this feature, because it is a little messy. The style embedding is given by `y_cond` [here](https://github.com/openai/jukebox/blob/2705dc07b29439b23fab5b4d6e85597ad7d90da3/jukebox/prior/prior.py#L232). To interpolate them, you could add these lines and pass...

Yeah, it's possible. You're welcome to try other variations

Huh, we've never seen this error before. It looks like we're using av version 8.0.2. Could you try a similar version? It'd be useful to make sure no audio files...

Thanks for catching `git checkout`. It's fixed now. Hm.. I haven't seen that before. What python/pip version are you using? Could you try again with python 3.7 and the latest...