Ditching OpenELM?

Has anyone fine-tuned Whisper for lyric transcription?

We've just released a notebook that shows how to add voice-consistency to Parler-TTS on mono-speaker datasets🚀🔥

How does it work? Simply create prompts that have a
consistent name (e.g. Jenny 🍀) and let the magic of fine-tuning work.

SPOILER: it does work fine on a free…

We need more open source music generation models. Udio, Suno, and now ElevenLabs are all closed source. Jukebox is older, restrictively licensed, and simply not as good. When will the new open source SOTA arrive?

Very cool! Also amazing to see the code + models released alongside the paper (instead of 'code coming soon')

Would you switch to OpenAI Search if it's a bit better than Google but significantly slower?

Looking for good large-scale sound effects/music/non-speech-audio datasets. Any suggestions?

Is it just me or does Apple's iPad ad look... somewhat AI generated?

github.com/mit-han-lab/qs…

Which one for the 3B variant?

Looking for a long-context (>16K) BERT-like model, any suggestions?

cc Alexander Doria

Welcome IBM Granite Code LLMs 🤗

> 34B, 20B, 8B & 3B models
> Base & Instruct - Apache 2.0 licensed
> Trained on 4.5T tokens using depth upscaling
> Covers 116 code languages ;)

Data:

> Uses The Stack for pre training
> Filters out low quality code
> Performs and exact and…

Looks like AK's account just got hacked

Why are Claude's input fields larger than their buttons? Noticed this in the iOS app too... is this a new design trend?

mrfakename It's a bit hacky so I'm not ready to put up a pr with proper doc cleanup BUT:
gist.github.com/kalomaze/661b7…
In a virtual env with Transformers installed you can hack this in by replacing the file in:
/venv/lib/python3.10/site-packages/transformers/models/mixtral/