Files
zeekayandHanzo Dev 6a363f7ee3 merge: upstream main ddf1b879 and the aarch64 repack branch 35d7ae7c
Three-way per file through the rename: base 39355c6c91, theirs the upstream
merge of main with 35d7ae7c (via 7ef6a694d), ours origin/main.

Lands the barrier thread pool, runtime CPU feature dispatch, aarch64 and x86
repacked quantized kernels (indexed_gemv, gemv_fused_shared_lhs), quantized
embedding on CPU/Metal/CUDA, paged flash-attn kernels, cutile fused MoE, the
in-place op aliasing API, parking_lot storage locks, Metal buffer builders and
debug labels, safetensors 0.8, hf-hub 1.0, tokenizers 0.23.

Ours keeps ROCm, Vulkan, wgpu, the IQ/MXFP4/NVFP4/TQ dtypes, and the i-quant
weight/activation block split in matmul. Our backends ride InplaceOpN and the
quantized embedding; Metal get_rows gains the IQ kernels. ug leaves, as upstream.

Co-authored-by: Hanzo Dev <[email protected]>
2026-09-15 15:52:28 -07:00
..
2026-01-16 12:30:55 -08:00

hanzo-ml-based

Experimental, not instruction-tuned small LLM from the Hazy Research group, combining local and linear attention layers.

Blogpost

Simple linear attention language models balance the recall-throughput tradeoff

Running an example

$ cargo run --example based --release -- --prompt "Flying monkeys are" --which 1b-50b --sample-len 100

Flying monkeys are a common sight in the wild, but they are also a threat to humans.

The new study, published today (July 31) in the journal Science Advances, shows that the monkeys are using their brains to solve the problem of how to get around the problem.

"We found that the monkeys were using a strategy called 'cognitive mapping' - they would use their brains to map out the route ahead," says lead author Dr. David J. Smith from the University of California