yu3zhou4

It's a very useful guide, I used it when working on torch-webgpu some time ago. Elie also published few packages that help with using some WebGPU related things, I don't remember exactly what was that but I recall it saved me lots of time

https://github.com/jmaczan/torch-webgpu

Print ondemand can be ordered for less than $5 per zine in Lulu, since 2022 or so. I guess Zach doesn’t make a cut from it since this $5 probably barely covers the cost of print and Lulu fee :(

PyTorch?

README is in my opinion (author here) the most interesting - I wrote it to help others build useful mental model to be able to recreate the project yourself, without need to even read my code

There was onivim that was a bit hyped a few years ago but unfortunately it died

Just learning math and trying myself in ml research

Laughed hard about Collegium Tumanum

Poland was sort of occupied until 1989

Good idea, too! Why do you see my explanation as cynical, though?

Maybe it’s in order to have an external provider to blame for failures and shift the blame/responsibility?

Adding a support for new hardware to PyTorch is actually quite convenient. I did that with WebGPU using the same PrivateUse1 mechanism TorchTPU used. Every hardware has its own slot and identifier, and when you want to add a support for a new one without merging it into PyTorch, PrivateUse1 works essentially like plug-in slot

https://github.com/jmaczan/torch-webgpu

They write that they use PrivateUse1, so it’s a custom out-of-tree backend

The system like iOS is still closed. Once Apple ends support for a device, being able to swap a battery won’t help much

Thx, great read about Yemen from Maciej Cegłowski https://idlewords.com/2014/07/sana_a.htm

An open course on building high performance LLM inference engine! Hope to finish by the end of April

https://github.com/jmaczan/tiny-vllm

I love this article, the edge cases are where the seeming “simplicity” of floating-point numbers breaks down

Recently wrote a chapter in tiny-vllm course about floats in context of LLM inference, much shorter and not that deep as this one, for anyone interested in topic you might like it too https://github.com/jmaczan/tiny-vllm?tab=readme-ov-file#how-...

Honestly there is a lot for room of improvement in torch-webgpu for performance. Needs involvement of community but the opportunities are definitely there

Overall I think it’s a good perspective and worth reading to learn your perspective and experience. Just for anyone who doesn’t come from a good place, please be kind to yourself. You probably really need to work exceptionally hard and need to have top 1% resume. And articles like this can be quite depressing. Don’t let it you down

The author tries to be mindful about it and I appreciate it, but from a perspective of someone from a shithole it still might feel bad

Referring to:

I've never gotten a job by applying to it. It's always been referrals or someone reaching out to me. So honestly, my resume is shit compared to my peers. I'm terrible at interviewing, and I've never done LeetCode. This is not a brag; it's just not my style. Am I a nepo baby? I don't know. Was I a morale hire? I'm pretty disagreeable. Was it merit? Also not sure

I recognize that not everyone has access to the same networking opportunities, and the traditional job application process can be a valid and necessary path for many. But social media, when used correctly, is a great way to get an opportunity

Good for the maintainer, hope they find peace and do things just for their fun, without needing to deal with comments like that anymore

So in some sense we start recreating an operating system, or at least the userspace, within the Claude code. There was some name for this pattern but I can’t recall

Can confirm that it also happens in other complex systems! Still a lot of good time and the novelty factor helps with pushing through

Congrats and happy for you, you had a lot of fun and did something genuinely interesting

I’m recreating a tiny version of vLLM in C++ and CUDA from scratch (high throughput LLM inference server)

https://github.com/jmaczan/tiny-vllm

Liar paradox

I had to use AI to actually understand what you wrote it and I think it's an underrated comment

Thank you very much!

Please share prompts or md files so we can generate more!

I try my best to push ML things into WebGPU and I think it has a future, but performance is not there yet. I have little experience with Vulkan except toy projects, but WebGPU and Vulkan seem very similar

PyTorch compiler and runtime for WebGPU!

ngmi