Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Federated finetuning of Whisper on Raspberry Pi 5

flower.dev

Post reply on HN
Print view View original

11–20 of 21 posts

Re: Federated finetuning of Whisper on Raspberry Pi 5

#11

Post by jafermarq »

danieljanes wrote: post #10
coder543 wrote: post #9

Earlier quoted context omitted.

v3 only comes in one flavor: large. I don’t think you’re going to have a good time running the large model on a Pi of any kind. The large models are 32x slower than the tiny models, roughly.[0] I just tested, and whisper.cpp on my Pi 4 can transcribe the 30-second a13.wav sample (“make samples” to fetch it) in 18.5 seconds. You can do the math… 32x = 10 minutes transcribe 30 seconds of audio with the large model. Not…

I can confirm that we're seeing 2x to 3x faster (RPi 4 vs RPi 5) in some of our early tests

yes. Finetuning a whisper model on a RPi 5 is ~2x faster than on the RPi 4. Other stages involving data pre-processing with HF dataset is again 2x-3x faster.

Re: Federated finetuning of Whisper on Raspberry Pi 5

#12

Post by jafermarq »

a_wild_dandan wrote: post #5
filterfiber wrote: post #2

I don't think the article mentions it, how well does the rpi 4 and 5 do for inference with whisper especially v3?

I’m also interested in peoples’ experience. I’d expect decent performance: Whisper 3 has many model sizes, down to 35Mb, iirc. Training, and especially inference, should be doable on a Pi5.

Yes. The example uses Whisper-tiny which is 39M, a perfect match for the downstream task of keyword spotting. Just one line needs to be changed in the code to run a larger Whisper model :)

Re: Federated finetuning of Whisper on Raspberry Pi 5

#13

Post by saqadri »

This is cool. This might be a silly question, but what are the scenarios where it's useful for fine-tuning on the edge with small devices? I get inference on the edge, and curious about metrics on that for Whisper, but isn't it better to fine-tune on beefier infrastructure and then deploy it for inference on the edge?

Re: Federated finetuning of Whisper on Raspberry Pi 5

#14

Post by kkielhofner »

a_wild_dandan wrote: post #5
filterfiber wrote: post #2

I don't think the article mentions it, how well does the rpi 4 and 5 do for inference with whisper especially v3?

I’m also interested in peoples’ experience. I’d expect decent performance: Whisper 3 has many model sizes, down to 35Mb, iirc. Training, and especially inference, should be doable on a Pi5.

> Whisper 3 has many model sizes

Nitpick but important - Whisper v2 and v3 are large only. It's actually the same Whisper but the version of the model (large-v2, large-v3) has been updated.

All of the other model sizes are the original release.

Re: Federated finetuning of Whisper on Raspberry Pi 5

#15

Post by danieljanes »

saqadri wrote: post #13

This is cool. This might be a silly question, but what are the scenarios where it's useful for fine-tuning on the edge with small devices? I get inference on the edge, and curious about metrics on that for Whisper, but isn't it better to fine-tune on beefier infrastructure and then deploy it for inference on the edge?

The big opportunity on the edge is access to more data. Especially with the rise of end-to-end encryption, applications will be able to use more (and more diverse) data on the edge to get better model performance. It's generally true that training on beefier infrastructure is easier, but in the long run, nothing can beat access to better data. And edge hardware has gotten a lot faster over the last few years.

Re: Federated finetuning of Whisper on Raspberry Pi 5

#16

Post by lfmunoz4 »

ulnarkressty wrote: post #4

How would this actually work in practice? Do I ask the user to utter specific words then train on that? How is it different from the traditional speech recognition that I need to 'train' to work better on my voice? The Holy Grail would be to train the model while using it, without any friction. I don't think these methods support that though.

The device on the edge creates the data but must also label it, right?

Re: Federated finetuning of Whisper on Raspberry Pi 5

#17

Post by triyambakam »

saqadri wrote: post #13

This is cool. This might be a silly question, but what are the scenarios where it's useful for fine-tuning on the edge with small devices? I get inference on the edge, and curious about metrics on that for Whisper, but isn't it better to fine-tune on beefier infrastructure and then deploy it for inference on the edge?

It seems like one benefit of fine tuning on the edge is the data doesn't need to move around as much. My father taught me "don't move a pile of dirt twice", so maybe it is like that.

Re: Federated finetuning of Whisper on Raspberry Pi 5

#18

Post by jafermarq »

lfmunoz4 wrote: post #16
ulnarkressty wrote: post #4

How would this actually work in practice? Do I ask the user to utter specific words then train on that? How is it different from the traditional speech recognition that I need to 'train' to work better on my voice? The Holy Grail would be to train the model while using it, without any friction. I don't think these methods support that though.

The device on the edge creates the data but must also label it, right?

If the task requires labels yes. Alternatively, an auxiliary model (not part of the training) can generate pseudo labels and use those during training. A more general approach (for which a few works in FL have proven to work pretty well) would do instead un/semi-supervised training.

Re: Federated finetuning of Whisper on Raspberry Pi 5

#19

Post by a_wild_dandan »

kkielhofner wrote: post #14
a_wild_dandan wrote: post #5

Earlier quoted context omitted.

I’m also interested in peoples’ experience. I’d expect decent performance: Whisper 3 has many model sizes, down to 35Mb, iirc. Training, and especially inference, should be doable on a Pi5.

> Whisper 3 has many model sizes Nitpick but important - Whisper v2 and v3 are large only. It's actually the same Whisper but the version of the model (large-v2, large-v3) has been updated. All of the other model sizes are the original release.

I reread your comment multiple times and still don’t understand the important nitpick. Are you saying that the smaller models haven’t been updated alongside the Whisper 3 release? That makes the most sense to me, but I don’t want to misinterpret what you mean!

Re: Federated finetuning of Whisper on Raspberry Pi 5

#20

Post by account17 »

a_wild_dandan wrote: post #19
kkielhofner wrote: post #14

Earlier quoted context omitted.

> Whisper 3 has many model sizes Nitpick but important - Whisper v2 and v3 are large only. It's actually the same Whisper but the version of the model (large-v2, large-v3) has been updated. All of the other model sizes are the original release.

I reread your comment multiple times and still don’t understand the important nitpick. Are you saying that the smaller models haven’t been updated alongside the Whisper 3 release? That makes the most sense to me, but I don’t want to misinterpret what you mean!

They only released the "large" model of both v2 and v3, the tiny model is v1
Post reply on HN