RahulBhalley

icon indicating a website link https://scholar.google.com/citations?user=5hIJB7oAAAAJ&hl=en icon indicating an email [email protected]

icon location Chaotic Latent Vector icon location I am interested in developing artificial intelligence software.

Results 122 comments of RahulBhalley

Due to recent development, maybe release training components/scripts?

I have access to an RTX 4090. Only problem is inference in production environments. This repo's Tortoise inference speed is sadly insanely slow!

Add transfer learning support to image classification models

It's been a long time. Any progress on this?

Export to JIT script

I can try to train it. But I have some queries: - How much memory and time can it take to train the model? - Can I mix languages such...

Export to JIT script

Thanks for the suggestions! > @RahulBhalley The author of the repo said he trained it on a single gpu(so something like a 3090?). In terms of time I'm not sure....

Export to JIT script

Look that this statement. https://github.com/lifeiteng/vall-e/blob/27c0667e0c3a18fd9074863a88a88094c6edbb2d/valle/bin/trainer.py#L148 It uses `ScaledAdam` (a custom implementation from scratch). https://github.com/lifeiteng/vall-e/blob/27c0667e0c3a18fd9074863a88a88094c6edbb2d/valle/modules/optim.py#L129 But scaled Adam doesn't exist in `bitsandbytes`. I don't know if using AdamW from `bitsandbytes` will...

Export to JIT script

> I see. I mean see how much vram you save. If it's only something like 3gb. Is it really worth? The point of using 8bit optimisers is mainly for...

Export to JIT script

Haha, okay.

MPS: models require an initial pass for reproducibility

Why GH closed it? It should remain open until solved.

MPS: models require an initial pass for reproducibility

Yeah bot. Address it.

MPS: models require an initial pass for reproducibility

We might instead close it as now Apple has released SD with CoreML.