And just like that, Nx 1.0 (along with Complex, EXLA and Torchx) is out!
Now on Track 2: 'Sharding and the Future of Numerical Elixir' with Paulo Valente (@polvalente).
#ElixirConf #ElixirConfUS #ElixirConfUS2026 #nx #mlView quoted post on X
Paulo Valente
@polvalente
Software Engineer
Maintainer of Nx, the Numerical Elixir ecosystem
Sponsors go toward further Nx development: github.com/sponsors/polva…
Maintainer of Nx, the Numerical Elixir ecosystem
Sponsors go toward further Nx development: github.com/sponsors/polva…
197 Following 908 Followers 1K Posts
Latest posts 5
Handrolled implementation is now at 320 tok/s for the 0.6B-4bit variant, while plain Axon rewrites can get the bumblebee model to 100 tok/s on the non-quantized variant!
New Hex version shipping tomorrow.
New Hex version shipping tomorrow.
In the past few days I shipped a proper EMLX compiler that leverages MLX tracing compiler, with support to all Nx constructs.View quoted post on X
This was inspired by the other MLX-based backend. PR below has Qwen3 perf comparison. We can now reach 170+ tok/s!
github.com/elixir-nx/emlx…
In the past few days I shipped a proper EMLX compiler that leverages MLX tracing compiler, with support to all Nx constructs.
This was inspired by the other MLX-based backend. PR below has Qwen3 perf comparison. We can now reach 170+ tok/s!
github.com/elixir-nx/emlx…
This was inspired by the other MLX-based backend. PR below has Qwen3 perf comparison. We can now reach 170+ tok/s!
github.com/elixir-nx/emlx…
github.com feat: lower while and runtime calls by polvalente · Pull Request #123 · elixir-nx/emlx This PR removes the need for graph splitting on runtime calls and while loops. Benchmarks now stand as follows: each entry in the table is row / col. > 1 means row is faster than col Qwen3... EMLX just got a nice boost: github.com/elixir-nx/emlx…
This PR introduces the EMLXAxon library which provides a few Axon rewriters based on EMLX Fast callbacks.
Local benchmarks show 4x perf on Bumblebee code and 6x if using a port of Qwen3-0.6B-4bit
0.3 soon!
This PR introduces the EMLXAxon library which provides a few Axon rewriters based on EMLX Fast callbacks.
Local benchmarks show 4x perf on Bumblebee code and 6x if using a port of Qwen3-0.6B-4bit
0.3 soon!
github.com/elixir-nx/emlx…
Quantization support just got merged! It still needs a bit of validation, but EMLX 0.3 is close!
Quantization support just got merged! It still needs a bit of validation, but EMLX 0.3 is close!
github.com feat: Add 4-bit quantization support for LLM inference on Apple Silicon by notactuallytreyanastasio... Summary This PR adds quantized tensor operations to EMLX, enabling efficient large language model inference on Apple Silicon GPUs. It powers a pure Elixir LLM inference stack achieving 135 tok/s on... This week EMLX got quite a bit of work done.View quoted post on X
github.com/elixir-nx/emlx
This includes: Updated to MLX 0.31.2, added support for CommandQueues for thread isolation,
native LinAlg operations, zero-copy to_binary, and more!
0.3 coming soon, right after quantization ships!