baracuda-transformer-engine
v0.0.1-alpha.86 ExperimentalSafe Rust wrapper for baracuda's port of NVIDIA TransformerEngine's FP8 cast/transpose + delayed-scaling recipe primitives. Provides `Fp8Recipe` (delayed-scaling state with amax history), `Fp8CastPlan` for {f32, f16, bf16} → FP8 with running amax, `Fp8DequantPlan` for FP8 → {f32, f16, bf16}. Cast/recipe subset only — `normalization` / `fused_rope` / `fused_attn` / `fused_softmax` / `activation` / `gemm` skipped (overlap existing baracuda phases). NO cuDNN dep, NO pybind11. On Ada (sm_89) the FP8 wins are bandwidth-saving only (KV cache, weights); FP8 tensor-core math throughput equals BF16. Forward-compatible with Hopper / Blackwell where the compute wins also materialize.
Quick Verdict
- ✓Actively maintained (updated 3d ago)
- !Pre-1.0: API may have breaking changes
- ✓Permissive license (MIT OR Apache-2.0)
Security
Deep Insights
62 downloads in the last 30 days, down 36% from the previous period. May indicate migration to alternatives.
The primary maintainer publishes 187 crates. This suggests deep Rust expertise and long-term commitment to the ecosystem.
At 20KB, baracuda-transformer-engine is lightweight. Small crate size correlates with focused, well-scoped functionality.
Health Breakdown
Recency, release consistency, active ratio
Yanked ratio, deps, size, maturity, features
Reverse deps, ownership, ecosystem
Downloads, momentum, growth trend
Docs, repo, license, metadata
Download Trend
Top Dependents
Most downloaded crates that depend on baracuda-transformer-engine