forward-pass

Forward Pass Learning and Inference Library, for neural networks and general intelligence, Signal Propagation (sigprop)

This is a minimal, from-scratch implementation of a scalar-valued automatic differentiation (autograd) engine and a small neural network library built on top of it. It follows the ideas behind Andrej Karpathy's micrograd, rebuilt line by line to actually understand how backpropagation works — not just call it.

200 lines Flash Attention (only forward pass) in CUDA.

Bumblebee, Axon, and Nx adapter layer for compiling Crucible tap plans into model runs, hooks, traces, and decode steering.

LSTM cell forward propagation implemented in Excel formulas across three linked worksheets.

Tap-plan and probe-contract library for selecting, bounding, and negotiating model-internal forward-pass observations.

NumPy-only CNN implementation - Implements conv2d, ReLU, maxpool2d, flatten, dense, softmax and backpropagating from scratch

Implementasi forward pass Multi-Layer Perceptron (MLP) 1 hidden layer dari nol menggunakan numpy murni (tanpa framework deep learning).

Bounded forward-pass trace schema and persistence helpers for Crucible signal captures, layer trajectories, and decode telemetry.

In this repository is a programm with a pretrained artificial neural network from Building AI. The network predicts the price of a cabin based on it's attributes.

Routing, gating, fusion, uncertainty, verifier, shared memory, and steering decision contracts over Crucible signal traces.

NumPy implementation of Word2Vec Skip-Gram with Negative Sampling. Trains word embeddings from raw text with manual forward pass, negative sampling loss, gradients, and SGD.

Canonical Elixir signal ontology for transformer forward-pass artifacts, tensor summaries, capabilities, and internal-control surfaces.