canyon289

7 ms·

1.
by canyon289 22d ago
A lot of different ways. Every guide on this guidebook runs on a ~$500 gpu or a couple of dollars on a cloud instance with a bigger GPU https://ravinkumar.com/GenAiGuidebook/book_intro.html This guidebook covers pretra
2.
by canyon289 1mo ago
Disclaimer, I work on Gemma and open models at Deepmind and the opinions here are my own There were open models from EleutherAI (GPT-Neo), Google Brain (T5X, Bert), and HuggingFace was promoting open models (and others doing open work I hav
3.
by canyon289 2mo ago
Join the waitlist! It's in the blogpost
4.
by canyon289 2mo ago
Totally fair and thank you for nit picking! I edited my comment. AI2 is doing amazing stuff, I read their papers and code all the time and appreciate how open they are.
5.
by canyon289 2mo ago
I'm a researcher at Deepmind that contributed to these models. (And the opinions here are my own) Just want to say, Deepmind is a great place to work and the only (Edit: one the few unique labs!) lab where you can move from large front
6.
by canyon289 5mo ago
This is a nice writeup!
7.
by canyon289 6mo ago
This is going to sound like a corp answer but I mean this genuinely as an individual engineer. Google is a leader in its field and that means we get to chart our own path and do what is best for research and for users. I personally strive t
8.
by canyon289 6mo ago
I dont have the metrics off hand, but I'd say try it and see if you're impressed! What matters at the end of the day is if its useful for your use cases and only you'll be able to assess that!
9.
by canyon289 6mo ago
On this one I dont know :) I'll ask my friends on the evaluation side of things how they do this
10.
by canyon289 6mo ago
Its hard to say because Pixel comes prepacked with a lot of models, not just ones that that are text output models. With the caveat that I'm not on the pixel team and I'm not building _all_ the models that are on google's dev
11.
by canyon289 6mo ago
We are always figuring out what parameter size makes sense. The decision is always a mix between how good we can make the models from a technical aspect, with how good they need to be to make all of you super excited to use them. And its a
12.
by canyon289 6mo ago
You could try Gemma4 :D
13.
by canyon289 6mo ago
Hi all! I work on the Gemma team, one of many as this one was a bigger effort given it was a mainline release. Happy to answer whatever questions I can
14.
by canyon289 7mo ago
I work on Gemma and Gemini models I want to echo Daniel's point here. Small finetuned models have their place even with larger general purpose models. For example last year with Daniel/Unsloth's help we released a tiny specia
15.
by canyon289 9mo ago
I'm with you! Small generative models are awesome, I thought so a decade ago and I still think so now! The size of what is "small" has definitely increased though, I used to think a 100 parameter model was large back in 2016,
16.
by canyon289 9mo ago
Good insight here, we actually did not include thinking into this model partly because we saw how incredibly fast it was to just get the minimum amount of tokens to output an answer. Thinking helps performance scores but we'll leave it
17.
by canyon289 9mo ago
Its definitely a step in that direction. I use Gemma models on my local macbook all the time and am personally excited to have this one available for me at home now as well
18.
by canyon289 9mo ago
It depends on a couple of things. If you expect reasoning or frontier level chat abilities then larger Gemma models or Gemini is better. Another hard constraint is context limit, Gemma 270m is at 32k so if the search results returned are ma
19.
by canyon289 9mo ago
I've only just skimmed this blog post but if I'm reading correctly FunctionGemma can work just like what's intended here, a "contextless" tool router. Going one level up you as a developer have a choice how much con
20.
by canyon289 9mo ago
I want to say so much right now but I can't :) The most generic thing I can say is I really do like working at Google because its one of the few (maybe only) company that has models of all sizes and capabilities. Because of this resear
21.
by canyon289 9mo ago
:popcorn gif:
22.
by canyon289 9mo ago
Hi, I'm not on the t5 Gemma team but work on gemma in general. Encoder Decoder comes from the original transformers implementation way back in 2017. If you look at figure 1 you'll see what the first transformer ever looked like. S
23.
by canyon289 9mo ago
I'm not specifically promising anything but I do want to say 2026 is going to be a great year! Many of my colleagues are shipping models too, such as t5gemma which is on the front page, and I'm personally excited to see what we&#x
24.
by canyon289 9mo ago
> Do you recommend any particular mix or focus in the dataset for finetuning this model, without losing too much generality? Astute questions, there's sort of two ways to think about finetuning, 1. Obliterate any general functionali
25.
by canyon289 9mo ago
We evaluate many things that you alluded to, such as speed on device, output correctness, and also "is this something that would be useful" the last one being a bit abstract. The way we think about it is what do we think developer
26.
by canyon289 9mo ago
> How good is the model at composing actions? I think you mean taking the results of one function call and putting it into another? We saw some promise but didn't heavily train for this use case in the base model. The thing we notic
27.
by canyon289 9mo ago
I just looked through their webpage and github and I'm not sure. But maybe someone should make a feature request! https://github.com/n8n-io/n8n
28.
by canyon289 9mo ago
Its already on the phone! Check out the demo videos and colab that show you how to load this model onto a device relatively easily. On this project I was lucky enough to work with the Google AI Edge team who have deep expertise in edge depl
29.
by canyon289 9mo ago
Yes a separate model is performing ASR in this case. Gemma270m (base, function, and others) are not multimodal out of the box. That being said if someone in the community wanted to use other encoders like siglip and plug them into Gemma270m
30.
by canyon289 9mo ago
Thank you much for the kind words

More ›