BS-RoFormer

BS-RoFormer copied to clipboard

MelRoformer parameters from paper

16

I'm trying to reproduce the paper model. But I have no luck. My current settings which gave me batch size only 2 for 48 GB memory: ``` dim: 192 depth:...

ZFTurbo

Feature request: decouple the loss function of the forward function

2

In the current implementation, the `forward()` method is generic for train or eval mode. In some case, we need to have not only the loss but the prediction on output...

dorpxam

After separating STFT and ISTFT from the `BSRoformer` class, I was able to successfully export the model to ONNX, and `trtexec` could convert the ONNX model to a TensorRT engine....

bfloat16

Open sourcing my Mel-Band Roformer vocal model

2

Hello @lucidrains I have just open sourced my Mel-band Roformer vocal model, it performs slightly better than the authors model because I trained it with more data. I also included...

KimberleyJensen

Hi! Amazing work! Any chance it can run on MPS (Metal Performance Shaders) - Apple GPU?

dj-nuo

Hello, Firstly, thank you for making the code publicly available. I am attempting to reproduce the results of the BS-RoFormer paper to achieve similar SDR performance and am referencing your...

EuiYeonKim

hello, I have two question about the Linear Attention that was added later. Can you clear me up why it is called Linear Attention when the referenced paper introduces Cross-Covariance...

Psarpei

Is it possible to reduce some aspects of Melband to give good results still but be able to train on smaller GPUs? I have tried finetuning Kimberley's model with my...

lucellent