numerical-precision

Here are 6 public repositories matching this topic...

vt-asaplab / ZIP

Zero-Knowledge AI Inference with High Precision (ACM CCS 2025)

python go rust machine-learning artificial-intelligence zero lookup-table verifiable-computation zero-knowledge-proofs ieee-754-floating-point aiaas plonk zkml numerical-precision caulk high-precision-approximation high-precision-zero-knowledge-proofs zero-knowledge-machine-learning

Updated Jan 22, 2026
C

Hiramdu / greedy-decoding-cross-precision-divergence

Star

Official code for the TMLR 2026 paper on BF16/FP16 divergence in LLM greedy decoding and selective FP32 lm_head recomputation.

cuda transformers pytorch reproducibility fp16 greedy-decoding bf16 fp32 large-language-models llm-inference tmlr numerical-precision

Updated Sep 22, 2026
Python

tanviranindo / posit-federated-learning

Star

Posit-Enhanced Docker-based Federated Learning: Bridging Numerical Precision and Cross-Architecture Deployment in IoT Systems

docker machine-learning distributed-system posit federated-learning cross-architecture numerical-precision

Updated Apr 7, 2026
Python

LUOaini1213 / fp16x3-transformer

Star

fp32-accurate Transformer layer on fp16 tensor cores: hi/lo operand splitting (fp16x3) with fused Triton kernels, 2.83x median over the fp32 reference on a free T4 at ~1e-6 error, and the 100,000-token shape running in 14 GB. Built for TikTok TechJam 2026 Track 3.

gpu cuda pytorch transformer triton tensor-cores numerical-precision