HackerTrans
TopNewTrendsCommentsPastAskShowJobs

crowwork

no profile record

Submissions

Modern GPU Programming for MLSys

mlc.ai
80 points·by crowwork·il y a 22 jours·25 comments

Building an Open ABI and FFI for ML Systems

tvm.apache.org
2 points·by crowwork·il y a 9 mois·0 comments

Open ABI and FFI for Machine Learning Systems

github.com
1 points·by crowwork·il y a 9 mois·1 comments

[untitled]

1 points·by crowwork·il y a 2 ans·0 comments

[untitled]

1 points·by crowwork·il y a 2 ans·0 comments

MLCEngine: Universal LLM Deployment to Both Cloud and Local Devices

blog.mlc.ai
2 points·by crowwork·il y a 2 ans·0 comments

Running Google's Gemma 2B on Android

old.reddit.com
2 points·by crowwork·il y a 2 ans·0 comments

MLC Chat: Chat with Open Language Models Locally on iPad and iPhone

apps.apple.com
2 points·by crowwork·il y a 3 ans·0 comments

WebLLM NPM Package

npmjs.com
1 points·by crowwork·il y a 3 ans·0 comments

Bringing Hardware Accelerated Language Models to Android Devices

github.com
2 points·by crowwork·il y a 3 ans·0 comments

Bringing Hardware Accelerated Language Models to Consumer Devices

mlc.ai
1 points·by crowwork·il y a 3 ans·0 comments

MLC LLM – Large Language Models on iPhone GPU and Many More GPU Platforms

mlc.ai
2 points·by crowwork·il y a 3 ans·0 comments

MLC LLM: Universal LLM Deployment with GPU Acceleration

github.com
3 points·by crowwork·il y a 3 ans·1 comments

Web LLM

github.com
6 points·by crowwork·il y a 3 ans·0 comments

Web Stable Diffusion

github.com
254 points·by crowwork·il y a 3 ans·41 comments

Running Stable Diffusion fully in browser with WebGPU

mlc.ai
42 points·by crowwork·il y a 3 ans·19 comments

comments

crowwork
·il y a 9 mois·discuss
The goal of the project is to bring open ABI and FFI for machine learning systems.

- Stable, minimal C ABI designed for kernels, DSLs, and runtime extensibility. - Zero-copy interop across PyTorch, JAX, and CuPy using DLPack protocol. - Compact value and call convention covering common data types for ultra low-overhead ML applications. - Multi-language support out of the box: Python, C++, and Rust (with a path towards more languages).
crowwork
·il y a 2 ans·discuss
Scale LLM serving with programmable cross-engine serving patterns, all in a few lines of Python
crowwork
·il y a 2 ans·discuss
XGrammar is an open-source library for efficient, flexible, and portable structured generation. Bring 2x-10x speedup in grammar grammar-guided(JSON and CFG) LLM serving.
crowwork
·il y a 2 ans·discuss
Comes with ability to do full structured generation with json schema

also a in-browser demo https://chat.webllm.ai/
crowwork
·il y a 2 ans·discuss
runs on qwen2 on iphone with 26 tok/sec and a OpenAI style swift API
crowwork
·il y a 2 ans·discuss
2b model running at 20tok/sec on iphone, nice potential for future applications
crowwork
·il y a 2 ans·discuss
Runs Phi-2 on Samsung S23 with pretty decent speed on Google Chrome browser.

LLM on browser on a phone
crowwork
·il y a 3 ans·discuss
You can also try out the vulkan backend, which we know should work for windows, although speed might be slower than rocm
crowwork
·il y a 3 ans·discuss
Yes, it works out of box and the blog contains a prebuilt python package that you can try out
crowwork
·il y a 3 ans·discuss
There is also vulkan support which should be more universal(also included in the post), for example, the post also shows running LLM on a steamdeck APU.
crowwork
·il y a 3 ans·discuss
Checkout the latest docs https://mlc.ai/mlc-llm/docs/ MLC started with demos and it evolved lately, with API integrations, documentations into an inference solution that everyone can reuse for universal deployments
crowwork
·il y a 3 ans·discuss
It certainly also involves generating code(e.g. WebGPU, vulkan) that are more akin to traditionally compiler, and more like graph and memory optimization. So indeed more than packaging.

Please checkout the course if you are interested
crowwork
·il y a 3 ans·discuss
There is a conda app that can be installed on macos
crowwork
·il y a 3 ans·discuss
You can try out the demo and benchmark yourself
crowwork
·il y a 3 ans·discuss
https://mlc.ai/mlc-llm/
crowwork
·il y a 3 ans·discuss
Supported platforms include:

- Metal GPUs on iPhone and Intel/ARM MacBooks

- AMD and NVIDIA GPUs via Vulkan on Windows and Linux

- NVIDIA GPUs via CUDA on Windows and Linux

- WebGPU on browsers (through companion project WebLLM).
crowwork
·il y a 3 ans·discuss
tvm runtime is pretty decent(~700k-2M level depending on dependency included), you can checkout tvm community and bring up the question there, i think there might be some common interest. There are impl of runtime for vulkan, metal that can be used as reference.
crowwork
·il y a 3 ans·discuss
I think instead what would be needed is a wgpu native runtime support for TVM. Like the implementations in tvm vulkan, then it will be naturally link to any runtime that provides webgpu.h

Then yah the llm_chat.js would be high-level logic that targets the tvm runtime, and can be implemented in any language that tvm runtime support(that includes, js, java, c++ rust etc).

Support webgpu native is an interesting direction. Feel free to open a thread in tvm discuss forum and perhaps there would be fun things to collaborate in OSS
crowwork
·il y a 3 ans·discuss
The WGSL are generated and compiled through TVM and embedded into the wasm.

I think what you mean is wgpu native support. At the moment the web gpu runtime dispatches to the js webgpu environment. Once TVM runtime comes with wgpu native support (like the current ones in vulkan or metal), then it is possible to leverage any wgpu native runtime like what Zig provide.

Additionally, currently tvm natively support targets like vulkan, metal directly which allows targeting these other platforms
crowwork
·il y a 3 ans·discuss
Webgpu will ship this year, so it will be more widely available pretty soon