Prince Canuma comments

Results 572 comments of


                                            Prince Canuma

BUG: Error msg during training - Timestamp must be non-decreasing for series attribute

Hi @stonelazy, Prince Canuma here, a Data Scientist at Neptune.ai, I want to personally inform you of the good news! This issue is now fixed on the latest release of...

BUG: Error msg during training - Timestamp must be non-decreasing for series attribute

Hi @stevenmanton Could you share a minimal reproducible code and the setup you have(i.e., no of GPUs and strategy) ?

BUG: Error msg during training - Timestamp must be non-decreasing for series attribute

Hi @mrembalski Bummer that you're experiencing some issues. No worries, we're here to help! I have a couple of questions that'll help me understand the situation better: - Are you...

BUG: Error msg during training - Timestamp must be non-decreasing for series attribute

Perfect! I am looking forward to it :)

Models should not need to be re-loaded between back-to-back prompts

Hey @neilmehta24 Thanks for reporting this! Could you share a reproducible example?

Models should not need to be re-loaded between back-to-back prompts

> The exceptions are different from one another between architectures, there is usually some state that is not being reset I usually use `stream_generate` in the manner you refer to...

Add IndexTTS

Awesome work @senstella! >@Blaizzy I just noticed IndexTTS uses non-standard BigVGAN, which incoperates conditioning inside of BigVGAN using an ECAPA-TDNN model, I agree with @lucasnewman 100% here. It's the approach...

Add IndexTTS

Thank you very much @senstella! I will review and merge it by Sunday 🚀 And will also check this breaking change.

Add IndexTTS

> Note that the test failure is related to https://github.com/Blaizzy/mlx-audio/issues/191, which is caused by mlx-lm's API refactoring. @senstella fixed in #194

Add IndexTTS

@senstella the biganv tests are failing, could you look at them?