ekzhang

7 ms·

1.
by ekzhang 8d ago
The spacing and style of the hand sketched lines are really perfect. Was it hard get such clean drawings, even using principal curvature fields?
2.
by ekzhang 2mo ago
Maybe try using https://www.npmjs.com/package/@jax-js/onnx ? It supports most ONNX ops and should be drop-in replacement, e.g., https://jax-js.com/d-fine Disclaimer: I am the author
3.
by ekzhang 8mo ago
I don’t think this is straightforward but it may be a skill issue on my part. It would require dockerizing headless Chrome with WebGPU support and dynamically injecting custom bundled JavaScript into the page, then extracting the results wi
4.
by ekzhang 8mo ago
Yes, we are actively working on it! The goal is to be a full ML research library, not just a model inference runtime. You can join the Discord to follow along
5.
by ekzhang 8mo ago
Hm okay, seems like an interesting set of benchmarks — let me know if there’s anything I can do to help make jax-js more compatible with your docker setup
6.
by ekzhang 8mo ago
Firefox doesn’t support WebGPU yet, you can run programs in the REPL through other backends like Wasm/WebGL: https://jax-js.com/repl See: https://caniuse.com/webgpu
7.
by ekzhang 8mo ago
Since ONNX is just a model data format, you can actually parse and run ONNX files in jax-js as well. Here’s an example of running DETR ResNet-50 from Xenova’s transformers.js checkpoint in jax-js https://jax-js.com/detr-resn
8.
by ekzhang 8mo ago
I don’t think tf.tidy() is a sound API under jvp/grad transformations, also it prevents you from using async which makes it incompatible with GPU backends (or blocks the page), a pretty big issue. https://github.com/ten
9.
by ekzhang 8mo ago
For sure! It looks like this is benchmarking the autodiff cpu time, not the actual kernels though, which (correct me if I’m wrong) isn’t really relevant for an ML library — it’s more for if you have a really complex scientific expression
10.
Show HN: Jax-JS, array library in JavaScript targeting WebGPU(ss.ekzhang.com)
84 points by ekzhang 8mo ago|22 comments
11.
Show HN: jax-js – JAX in pure JavaScript(github.com)
2 points by ekzhang 9mo ago|0 comments
12.
by ekzhang 9mo ago
Wanted to share this library that I'm open-sourcing now after working on it for the past year. You write code like in JAX/NumPy, but it’s fully interactive on the frontend and compiles down to shaders on the user’s GPU (with WebGP
13.
Show HN: jax-js, an ML library and compiler for the web(ekzhang.substack.com)
4 points by ekzhang 9mo ago|1 comments
14.
by ekzhang 1y ago
Thanks! I think that's a great idea. Containers are nice, but there's nothing quite like a "real" VM system in terms of learning + developing, and they can be very cheap if you're OK with the performance hit from ne
15.
Show HN: SSH-hypervisor – like SSH, but each user gets their own microVM(ekzhang.substack.com)
34 points by ekzhang 1y ago|4 comments
16.
ChickadeeOS, a teaching operating system for Harvard's CS 161(github.com)
2 points by ekzhang 1y ago|1 comments
17.
Building a cloud GPU notebook that boots in seconds(modal.com)
1 points by ekzhang 1y ago|0 comments
18.
by ekzhang 1y ago
Hello! Thrilled to be sharing something I've worked on for the past 8 months at Modal — a cloud, GPU notebooks product. We wanted to improve the Jupyter Notebook experience while offering high-performance, cloud compute. So we sought o
19.
Modal Notebooks, a real-time collaborative notebook with cloud GPUs(modal.com)
4 points by ekzhang 1y ago|1 comments
20.
How the jax.jit() JIT compiler works in Jax-JS(ekzhang.substack.com)
2 points by ekzhang 1y ago|0 comments
21.
by ekzhang 1y ago
Hello - I wanted to share a bit about an open-source project I’ve been working on, building a fundamental library for ML and numerical computing in the browser. There are different tradeoffs here, but I’d like to replicate the speed and erg
22.
How the jax.jit() JIT compiler works in jax-js(substack.com)
5 points by ekzhang 1y ago|1 comments
23.
How I Created the Amazon Logo(bileshendry.com)
6 points by ekzhang 1y ago|0 comments
24.
Realtime MLS-MPM particle system in Three.js WebGPURenderer(github.com)
2 points by ekzhang 1y ago|0 comments
25.
Modal SDKs for JavaScript and Go(modal.com)
2 points by ekzhang 1y ago|1 comments
26.
by ekzhang 2y ago
Just want to mention that these standalone python builds have been super important for simplifying a lot of dev tooling, even if users may not see it directly — I work for a cloud infrastructure company, and they’re what allow us to give us
27.
by ekzhang 2y ago
Sorry, just realized the misunderstanding. To clarify Modal still uses the kernel WireGuard module. The userspace part that’s in Go and not in other languages that we use is the wgctrl library.
28.
by ekzhang 2y ago
This is it. I like Rust a lot, but you gotta pick the right tool for the job sometimes
29.
by ekzhang 2y ago
Right, vprox servers act as multiplexed NAT instances with a VPN attached. You do still need the VPN part though since our containers run around the world, in multiple regions and availability zones. Setting the gateway to a machine running
30.
by ekzhang 2y ago
We have a native OIDC integration at Modal, as well! Every container gets a token. https://modal.com/docs/guide/oidc-integration

More ›