David Rubinstein

I'm pretty sure most of the current operations can be ported over pretty easily if the new `Tensor` class supports: - generate - elementwise - combine `row` and `column` of...

Hi @VijayIyer could you give an example of a file that didn't work and your development environment (OS, what you have installed etc.)?

I think we should leave it as is and add a comment to the README about which commit should be used to match the paper i.e. the one with the...

Hi @martingasser ! Good find. Feel free to PR a change or I will when I have time to update the code. Regarding releasing the code for training, we intend...

Hi @G-haoyu Good idea, and similar to what we do [in the typescript version](https://github.com/spotify/basic-pitch-ts/blob/main/src/inference.ts#L160), but I think there are some changes to this implementation that would better fix the issue...

I forget what the current interface looks like, but maybe it'd be better if the user is responsible for giving a callback that handles finding the file and downloading it...

You could write a bunch of callbacks and the user specifies which one they want to use. That way they can use the file one or maybe a GCS one...

An alternative to writing an in-house library is to use [cloudstorage](https://pypi.org/project/cloudstorage/) or [libcloud](https://libcloud.apache.org/) which seems to fit this use case.

Hi @davidNHK and @fheyen. We can't currently open source basicpitch.io as it contains some Spotify-internal React components that are not open sourced yet. The website largely uses a similar structure...

Hi. What version of keras are you using? Is it one different than specified in basic pitch's requirements? [StackOverflow](https://stackoverflow.com/questions/66178738/attributeerror-module-tensorflow-python-training-experimental-mixed-precision) suggests an incorrect version is at fault here