Neith GPU-compute ML inference & tensor ops — tensor primitive, GEMM/Conv2D/DepthwiseConv/attention, fused activations, LayerNorm/RMSNorm/BatchNorm, softmax/logsumexp, flash-attention, INT8/INT4/FP8 quantisation, GGUF importer, graph fusion optimiser, op profiling, autotune, determinism toggle, training gradients, ring-allreduce, runtime bridge
At a glance#
- Path:
libs/neith/compute-ml - Domain: neith (active)
- Owner: @GreyChimp
- Docs: deep-dive · systems one-pager
- Nx project:
@neith/compute-ml
Commands#
bash
pnpm nx test @neith/compute-ml
pnpm nx lint @neith/compute-ml
<!-- Generated starting point (audit F-3): owners, replace this note with usage examples and API notes. -->