Anemll

@anemll

ANEMLL (pronounced like "animal") Artificial Neural Engine Machine Learning Library, Open Source Project

Seattle, WA

Joined January 2025

349Following

489Followers

Pinned

Anemll@anemll · Jul 21

If every post-2020 Apple device lit up its Neural Engine at once, humanity would have ~20 zettainteger-ops-per-second of on-device AI oomph—about five times the cumulative floating-point tensor capacity of all NVIDIA GPUs sold in the same period. In practice, Nvidia’s…

18.0K

Anemll@anemll · Jul 24

MLP from dequantized int8 with CoreML 9.x test / example source code gist.github.com/Anemll/a9838e1…

636

Anemll@anemll · Jul 23

I think it should be CoreML 26 🤣

SStephen Panaro@flat · Jul 23

Incoming new coremltools looks like it has some nice bits: - 8 bit input/output tensors (previously all 8bit compute was kept internal) - >1 input can be enumerated shapes (👀ANE)

321

Anemll Retweeted

Awni Hannun@awnihannun · Jul 22

Sure, here's a sample run: Prompt: 3135 tokens, 249.467 tokens-per-sec Generation: 100 tokens, 24.264 tokens-per-sec Peak memory: 272.008 GB

3.0K

Anemll@anemll · Jul 20

It’s amazing how CoreML visualization looks like circuit / FPGA design ! Here is AltUp representation for Gemma 3n. It would be cool to see/edit a graph representation alongside PyTorch or MLX code during development or debugging 🤔

anemll's tweet image. It’s amazing how CoreML visualization looks like circuit / FPGA design ! Here is AltUp representation for Gemma 3n.
It would be cool to see/edit a graph representation alongside PyTorch or MLX code during development or debugging 🤔

6.0K