ioedward

7 ms·

1.
by ioedward 3y ago
AMD didn't submit results, despite geohot/Lisa Su's plan to get AMD on MLPerf: https://twitter.com/realGeorgeHotz/status/166980346408248934...
2.
by ioedward 3y ago
8 H100s would have enough VRAM to finetune a 70B model.
3.
by ioedward 3y ago
Not at all, PyTorch uses sub-classing (nn.Module) and other Pythonic features heavily that are not supported with Mojo.
4.
by ioedward 3y ago
Trying the fuse feature, seemed the most interesting: > spaghetti ≈ trumpet > Both spaghetti and a trumpet can be difficult to eat without making a mess—spaghetti with its long, slippery noodles, and a trumpet with its wide, flared be
5.
by ioedward 3y ago
Related: 1 year represented in a photo. https://eirikso.com/2011/01/04/one-year-in-one-image/
6.
by ioedward 3y ago
The technical term is GPGPU, general-purpose computing on GPUs, but I like to call them GPUs for short.
7.
by ioedward 3y ago
Nvidia's enterprise GPUs are surprisingly unreliable. Working on a 128 GPU A100 cluster on AWS, 1 would fail every few days. I didn't have any insight on whether it was a hardware or software failure.
8.
by ioedward 3y ago
You also need the optimizer (e.g. Adam)'s state, which is usually double the parameter's size. So if using fp16, one parameter takes up 6 bytes in memory.
9.
by ioedward 3y ago
TPUs are mostly hoarded by Google Research (including Deepmind) and Ads. Very few are being used by external people.
10.
by ioedward 3y ago
The twitter thread is worth reading, Yann responds to some of the questions raised here.
11.
by ioedward 3y ago
No it won't. Large language models are trained on 1,000 - 50,000 GPUs. No one's going to buy hundreds of Mac pros to mount them in a datacenter for training ML models.
12.
by ioedward 3y ago
Normally people split up the model across multiple GPUs, i.e. model/tensor parallelism.
13.
by ioedward 4y ago
Doesn't seem to be affiliated with OpenAI. Domain was registered in 2017, 2 years after OpenAI was created. Trolling/domain squatting?
14.
by ioedward 4y ago
> I understand that PyTorch is an awesome tool for researchers, but it doesn't necessarily fit into a prod environment. Some of the largest deployments of ML are using PyTorch models, e.g. OpenAI, Meta, Microsoft.
15.
by ioedward 4y ago
Where did you buy the Instinct hardware and what's a good price?
16.
by ioedward 12y ago
Archive.org: https://web.archive.org/web/20150120162544/http://www.sbb.co...
17.
by ioedward 12y ago
So are you still shipping Python/Django as an executable with atom-shell? Or is this purely client-side javascript?
18.
by ioedward 12y ago
I'm getting stuck on the sign in page. After authorizing the app with Github, there's a black box that appears and the Sign In button shows up again. Edit: Looks like the server is 500ing, according to the Javascript console.
19.
by ioedward 12y ago
dang killed it because he didn't like it. (I'll go look for the source) edit: source - https://news.ycombinator.com/item?id=7693262 > Ok, when people start racing to post these at midnight, and beg for upvotes
20.
by ioedward 12y ago
Would love to hear more about the terminal as well - how did you input commands to it, and how does it compare to using a phone or computer?
21.
by ioedward 12y ago
I emailed you, but also wanted to express in the comments that I would be definitely interested in idempotent json patches.
22.
by ioedward 12y ago
Sure! I would _not_ advise getting the Nexus 5. As noted below in other comments, the image quality and autofocusing speed has much to be desired. The Camera2 API that supports DNG is only available on Android 5.0 phones. http://
23.
by ioedward 12y ago
There's nothing technical that's stopping them from doing it. I expect third party apps to do something similar once Lollipop has a higher market share.
24.
by ioedward 12y ago
All hand held, shot from 1/35th of a second to 1/1000th of a second. HDR Pro does some basic auto aligning and ghost removal, AFAIK.
25.
by ioedward 12y ago
I've been messing around with the Google Camera app and the new Camera API provided by Android 5. It turns out that you can actually obtain a lot higher quality images by using the DNG (digital negative) of the photo, instead of using
26.
by ioedward 12y ago
Okay, I misunderstood you then. I haven't found any words ending with "io" that are still available.
27.
by ioedward 12y ago
No real words? I just ran my script through a list of the most common english words, and it returned 4k .IO domains. I'll post them if anyone's interested.
28.
Show HN: All available three letter .io domains(edward.io)
13 points by ioedward 12y ago|11 comments