Searching...
10 results for “execution speed”
and and you use the same model instead of a bigger model with more parameters. You use the same model kind of recurrently. You you make it think for a while, and then and then you you optimize that sort of the thinking trace with, with reinforcement
34:20 / 35:36
from transcript34:20 – 35:36
models and and plus processors. Yep. I'm just curious if you've obviously, you've thought about it, but, like, what's your commentary? Yeah. I mean, I think there's a bunch of interesting trends. So energy based models is one, you know, diffusion bas
39:44 / 40:49
from transcript39:44 – 40:49
both the test time and and and when you're training the this reasoning system, you get more performance. You get I think it's about, like, quadratically more performance. It's like it's like I think you you if you put in 10 x more compute, you get, a
35:12 / 36:17
from transcript35:12 – 36:17
quite nice about the flash model is not only is it more affordable, it's also a lower latency. And I think latency is actually a pretty important characteristic for these models because we're going to want models to do much more complicated things th
8:14 / 9:19
from transcript8:14 – 9:19
Shopify was already running on the initial version of Ruby that was, I think, 10 times slower than what we have today if you look back upon the Ruby one eight six that Topi probably started on, just as I started on. And that was enough to propel Shop
1:27:43 / 1:28:54
from transcript1:27:43 – 1:28:54
that I don't know if it's launched or not yet. Yeah. I'm assuming I want to bring back to you know what you mentioned about, like, the as your guest, like, from Anthropic mentioned about the performance, improvement. I think there's a big change. I t
1:05:52 / 1:07:16
from transcript1:05:52 – 1:07:16
Cool. And then so then we'll it was the first world's fair. You talked about you'd we probably don't need fine tuning as as a fine tuning founder. Basically, I think your your talks are really good. I would recommend people watch all of them. What I
11:36 / 13:18
from transcript11:36 – 13:18
...and and speed, basically. And we we have models that completely define the frontier. So whatever your trade off is that you want as an individual user or as a as a developer, you should find one of our models satisfied as that constraint.
1:36:38 / 1:37:51
from transcript1:36:38 – 1:37:51
If your Rails request takes a second and a half to execute, the user's gonna notice. Your app is gonna feel slow and sluggish. You have to get that response time down below, let's say, at least three hundred milliseconds. I like to target a hundred m
1:28:25 / 1:29:31
from transcript1:28:25 – 1:29:31
So so when all of it route routes to one part of the model, then you can have the you can have this overloading of a sort certain set of the GPU resources or a certain set of the GPUs, and then the rest of the the training network sits idle because a
48:39 / 49:55
from transcript48:39 – 49:55
Have a podcast?
Run your podcast through Clypt
Get ranked clips, hooks, and ready-to-post copy from your own episodes. Free to try.