[New Pipeline] Add Grad TTS

[Vaibhavs10]

Note: This is still WIP

So far I've added all the text and model utils to two separate files in the GradTTS pipeline folder. I've also added the LinearAttention function in the attention.py file.

I am confused about adding GradTTS-specific Resnet blocks to resnet.py or directly using the ResNet2D block.

Next, I'll focus on adding the model and then the pipeline, any suggestions/ feedback is welcome.

cc: @patil-suraj & @patrickvonplaten

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint.

[patrickvonplaten]

Hey @Vaibhavs10 ,

Do you need a review here?

[Vaibhavs10]

Hi @patrickvonplaten - I am struggling a bit about splitting the Resnet blocks into a diffusers friendly format. So any quick pointers on that would be pretty helpful. after that, can we do a full review?

[patrickvonplaten]

Hey @Vaibhavs10 :-)

Yes I think grad-tts had some weird unet/resnet blocks. Could you for now just add something that works from where we can start iterating? Essentially try to build a very hacky block in unet_block and a completely new ResnetBlock (I think we should actually create a new resnet block here)

Once that's in the PR, I'm more than happy to help :-)

[Vaibhavs10]

Hey hey Patrick & team - sorry for the delay on this. Got pulled in different directions with work stuff. This is next on my priority list and will have some updates on this in the next couple days!

[patrickvonplaten]

Cool!

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.