Fine tune and Serve Faster Whisper Turbo

OTHER RESOURCES:

TIMESTAMPS:
0:00 Whisper Turbo Fine-tuning and Serving
0:39 Colab Demo: Transcribing Audio Files and Youtube Audio
2:34 How does Whisper (Turbo) work?
6:00 Faster Whisper, Insanely Fast Whisper, and Fast Whisper Server?
8:01 Fine-tuning Whisper Turbo for new words or accents
16:27 Automating training data cleanup with LLMs
18:36 Chunking our input audio and text data and pushing to hub
22:27 LoRA and Trainer Setup
25:15 Saving, evaluating and converting the model for OpenAI format and Faster Whisper
29:24 Setting up a Faster Whisper Server Endpoint

[savanthtadepalli]

Superb content as usual! You are a God Ronan!

Can you please make a video on a multi-agent use case that involves all three modalities and open source models. I would love to pay for such an end-to-end real world use case and learn from you! Please make it production grade.

[grtbigtreehugger]

Many thanks! I am working on transcription, so much appreciated!

[btaranto]

Very well structured your videos! From beginning to end! Thank you so much!

[iasadkamran]

I rarely comment or subscribe. You deserve a lot more than that 🎩🎩

[BhavikSonkusare]

very informative and useful video

Please make a video on how to use whisper in real time ASR.

[uwegenosdude]

Really cool stuff. Thanks a lot. BTW: We tried insanely fast whisper and it worked pretty well and it is really fast.

[hgwvandam]

Oh thanks man!, just what I was looking for!

[johnkidsman]

It would be perfect if you could have explained the data collator part in more detail! It's difficult to understand how the paddings work internally

[stepanda]

On 11:53 After connecting to Jupyter Notebook, I don't see Whisper fine-tuning notebook. Where should I get it and how to put it there? I've purchased Lifetime Access, but didn't receive any access details yet.

[thomashuynh]

Can you try fine-tuning on the lagre-v3 model and test fine-tuned model works as expected.