zcbenz
Joined 12 years ago
Karma
686
Submissions
77
Member since
Feb 27
https://zcbenz.com
View on Hacker NewsRecent submissions
Comment · 9 months ago
It is a bug in MLX that has been fixed a few days ago:
https://github.com/ml-explore/mlx/pull/3083
Comment · 1 year ago
In the absence of hardware unified memory, CUDA will automatically copy data between CPU/GPU when there are page faults.
Comment · 1 year ago
Most features are already supported on x86 CPUs, you can pip install mlx on Linux , and you can even use it on Windows (no official binary release yet but it is building and tests are passing).
Comment · 1 year ago
For learning internals of ML frameworks I recommend reading the source code of MLX: https://github.com/ml-explore/mlx .
It is a modern and clean codebase without legacies, and I could understand most things without seeking external articles.
Comment · 2 years ago
You can use any serialization that supports numbers, in the repo I used BSER (a binary format of JSON), you can even use raw JSON but it would be very inefficient for float numbers though.
Comment · 2 years ago
No it is not normal, I only tested x64/arm64 macs, I will try on linux.
Comment · 2 years ago
That is sad, the name sisi comes from the sisi empress:
https://en.m.wikipedia.org/wiki/Empress_Elisabeth_of_Austria
Comment · 2 years ago
A few months ago I ported ~15k lines of python code (10k are tests) to typescript, using GPT4. It cost me ~$70.
The python project is https://github.com/ml-explore/mlx and the converted project is https://github.com/frost-beta/node-mlx
I wrote a long prompt: https://github.com/frost-beta/node-mlx/blob/main/tests/promp...
The first result was almost always bad, but after manually modifying the assistant's answer, following generation usually went much better.