Parcollet Titouan

icon indicating a website link www.darnault-parcollet.fr icon indicating an email [email protected]

icon company Samsung AI Cambridge icon location United-Kingdom icon location Research Scientist at Samsung AI Cambridge — Affiliated Lecturer at the University of Cambridge — Associate Professor on leave at the University of Avignon

Results 231 comments of Parcollet Titouan

Real-Time ASR with Device Capture using torchaudio StreamReader

Hi! We have very limited bandwidth but we keep this proposal in mind!!

25,000 of diverse English ASR data (dataset name hidden) (code from Samsung AI Center Cambridge)

Here i'd love to have an opinion about having this kind of recipe into SB from @Adel-Moumen @pplantinga and @mravanelli . It's quite uncommon to have such a big part...

25,000 of diverse English ASR data (dataset name hidden) (code from Samsung AI Center Cambridge)

Agreed @pplantinga . Let me answer for the recipe part. All the scripts are different as the csv rows and steps / filtering are not the same. There is also...

25,000 of diverse English ASR data (dataset name hidden) (code from Samsung AI Center Cambridge)

Yes, but not now.

SpeechBrain and ONNX

I think that quite a few models could actually be compiled, for a long time we made sure that most models where jittable, which should be equivalent. I think it...

SpeechLLM (with LLaMA) and Conformer recipe for speech translation on CoVoST (Code from Samsung AI Center Cambridge)

I could use a first quick review on looking briefly at the code (forget about llama.py as it's from another PR). Maybe @Adel-Moumen / @pplantinga ? @poonehmousavi this is an...

Fix issue with weight loading for torch 2.6

Tbh, I couldn't reproduce the issue either, but it happened to our user. Isn't it ok to just merge this? It shouldn't break things for the people not affected, right?

SpeechLLM and Whisper

Hi @avishaiElmakies this sounds great! The first thing to determine is how much time you can devote to do things, then we can plan accordingly. @Adel-Moumen can add you to...

ASR training from scratch

Hello @nata-kostina, this is a template, which is expected to help learn SpeechBrain but very importantly, **it is not used to obtain a good ASR system**. That being said, it...

Replace deprecated torch.cuda.amp.custom_fwd with torch.amp.custom_fw

@Adel-Moumen should we merge without what you mentioned?