Pinned Loading

1. cute-ampere-dgemm cute-ampere-dgemm Public

   A DGEMM implementation in CuTe c++ targeting the ampere architecture

   Cuda

2. ampere-L2-partitions-microbenchmarks ampere-L2-partitions-microbenchmarks Public

   Replication of the experiments showcased by Citadel at GTC 2021 about the A100's L2 partitions.

   Cuda

3. ppo-jax ppo-jax Public

   a fully jitted & multi accelerator JAX implementation of PPO for discrete and continuous envs

   Python

4. Coral Coral Public

   SSL robot control software

   Rust 1

5. Llama 3.2 inference in mlx - functio... Llama 3.2 inference in mlx - functional style / no nn.Module

   1

   # Uses entropix's tokenizer & downloaded weights. Saving this file into entropix's repo should work

   2

   3

   import math

   4

   from typing import NamedTuple, List

   5

   from pathlib import Path