search · repositories, files, commits, branches, READMEs
Results for “metal”
Index built just now · 180 repositories · 86,397 paths
Repositories 1
-
forge
C++
Forge — LLM training from scratch in pure C++20 + Metal on Apple Silicon.
Files 20
-
src/core/metal_impl.cpp
forge · main
-
paper/sections/metal.tex
wp40_uqo · main
-
content/tools/metal.json
tools-spboucher-ai · main
-
src/ops/metal/metal_ops.h
forge · main
-
src/ops/metal/metal_ops.cpp
forge · main
-
docs/screenshots/metal-2m.png
metrika · main
-
data/enrich_cache/metalico.ca.json
fabri-ka · main
-
backends/metal/host/metal_backend.mm
wp40_uqo · main
-
third_party/metal-cpp/Metal/Metal.hpp
forge · main
-
Metrika/Sources/MetalScatterView.swift
metrika · main
-
third_party/metal-cpp/MetalFX/MetalFX.hpp
forge · main
-
tests/fixtures/edgar/metalinks_CIK0001620533_000162053326000018.json.gz
hfmarketdata · main
-
src/kernels/moe.metal
forge · main
-
src/kernels/adamw.metal
forge · main
-
src/kernels/quant.metal
forge · main
-
src/kernels/matmul.metal
forge · main
-
src/kernels/softmax.metal
forge · main
-
src/kernels/attention.metal
forge · main
-
src/kernels/embedding.metal
forge · main
-
src/kernels/layernorm.metal
forge · main
Commits 3
-
WP40 RAID: E1–E6 results (real 2024 panel, daily 2015–2024 histories), Parquet reader projection + fast paths, Metal GARCH per-series scaling, GARCH restarts, paper skeleton (intro, literature, references)
wp40_uqo ·
a9731b7· Simon-Pierre Boucher · 6 days ago -
WP40 RAID: Phase 3 — Metal backend (double-float accumulation, fused realized kernel, rolling sums, regression stats, batched GARCH NLL), accuracy/determinism tests, primitives microbenchmark + analysis script
wp40_uqo ·
5aa6672· Simon-Pierre Boucher · 6 days ago -
[ka6] fix connecteur niobec: frontal ADP intermittent PAR CID (4e cas, garde-fou retry borné répliqué) + correction d'un angle mort de culture — FORCE_CITY publiait l'offre talc de Timmins (ON) comme Saint-Honoré QC → bascule LANG en_US, found=0 = reflet exact de la source
job-ka ·
61a2ddf8· Simon-Pierre Boucher · 9 days ago
README content 10
-
forge
🔥 Forge LLM training from scratch — pure C++20 + Metal on Apple Silicon No PyTorch. No MLX. No ML dependencies. Just C++, Metal kernels, and a JSON parser. 📄 Paper · 🔬 Research notes · 📐 Spec ✨ What this is A complet…
-
wp40_uqo
⚡ RAID — Research Architecture for Intraday Data UQO Working Paper No. 40 — A Metal Accelerated C++ Framework for Reproducible High Frequency Financial Econometrics b31b1b?logo=latex&logoColor=white) 000000?logo=apple&lo…
-
wp33_uqo
…ed hypotheses (causal prefetch, topology aware placement, hierarchical caching, Metal kernels, adaptive orchestration on the latency–throughput–energy frontier), with an exact semantics constraint: placement decides spee…
-
llm-api
…oads. Two runtimes : MLX / MLX LM (preferred, safetensors) and llama.cpp (GGUF, Metal). Each model runs in its own worker process on — killing the process is how memory is guaranteed to come back. Memory policy : weights…
-
localvm-research
…d); 4 KiB QD1 random = 67 MB/s → 1 MiB QD8 = 13.8 GB/s (200× spread); saturated Metal GPU costs < 5% of SSD throughput 2 Margins predict disagreement (expG) AUROC 0.85–0.98 (3/4/8 bit); domain uniform; escalation for 99%…
-
phd_thesis
…s. swap dealers) conditions the reaction of energy (crude oil, natural gas) and metals (gold, silver, copper, palladium) futures to standardized macro surprises. Key findings 🎯 🧯 Increased speculative trading dampens t…
-
metrika
…PU (LAPACK) or GPU (MLX) automatically. Large bootstrap runs execute as batched Metal solves; you never choose a backend. Reproducibility is a feature, not an accident. All randomness flows through a counter based Philox…
-
prime-mystery-engine
…nodes, 162 CPU cores 5 min Cycle 4 10¹³ 3 nodes: 58 CPU cores + 3 GPUs (custom Metal kernel) 50 min The distributed design uses mergeable streaming statistics (exact integers < 2⁵³ in float64 → zero rounding , bit ident…
-
forge-studio
…on · zero third party dependencies ✨ What it is Forge is a from scratch C++20 + Metal LLM training framework: its own tensors, autograd, flash attention, Muon optimizer, MoE, QAT, and the zero copy weight format. Forge S…
-
zyquo-local
…is never required. Two prerequisites beyond the Command Line Tools: 1. Apple's Metal Toolchain on — MLX compiles its GPU kernels at build time. On a machine with Xcode 26+: ; a standalone on also works. 2. SDK pin on ma…