Files
Laurent Mazare 27996a1a9e Remove the old MFA gemm kernels. (#2742)
* Remove the old MFA gemm kernels.

* Use bf16 in helium on metal.
2025-01-26 20:36:31 +01:00
..
2025-01-13 18:04:14 +01:00

candle-helium: 2b LLM with CC-BY licensed weights

Helium-1 is a lightweight model with around 2B parameters, the preview version currently supports 6 languages, showing strong capabilities in those languages compared to existing open weights models.

Running the example

$ cargo run --example helium --release --features cuda -- --prompt 'Write helloworld code in Rust' --sample-len 150