krishanth7/everyo

How krishanth7/everyo works

EveryO is a CPU-first neural-computing framework for building and training small neural networks, evaluating them, and exporting supported models. Its documented learning loop runs from data and tensors through differentiable operations and model layers to loss, backpropagation, and optimizer updates. The graph also preserves documented data, inference, serialization, acceleration, interoperability, and experiment capabilities. Sampled trainer and operation references support internal links; several capability branches are represented by their exact tree paths, but their internal calls were not sampled.

Main entry points: trainer.py Training Loop · operations.py Differentiable Operations · layers.py Neural Network Layers · tensor.py Tensors and Autograd

Components 27

Tensor Core

Tensors and Autograd everyo/core/tensor.py
Differentiable Operations everyo/core/operations.py
Gradient Engine everyo/core/autograd.py
NumPy Backend everyo/backends/numpy_backend.py

Models and Training

Neural Network Layers everyo/nn/layers.py
Embedding Layer everyo/nn/embedding.py
Loss Functions everyo/nn/losses.py
Optimizers everyo/optim/optimizer.py
Training Loop everyo/training/trainer.py
Inference everyo/inference.py

Data and Evaluation

Data Preparation everyo/data
Datasets everyo/datasets
Dataset Abstractions everyo/data/dataset.py
Batch Loading everyo/data/dataloader.py
Evaluation everyo/evaluation.py

Runtime Integrations

Save and Load everyo/serialization
ONNX Export everyo/serialization/onnx_export.py
Dynamic Quantization everyo/quantization.py
CUDA Operations everyo/cuda/interface.py
CUDA Bindings cuda/bindings/bindings.cpp
TensorFlow Reference everyo/backends/tensorflow_backend.py

Experiments and Assurance

Experiment Tools everyo/experiments.py
Experiment Tracking everyo/experiment_tracking.py
Monitoring everyo/monitoring.py
Validation and Verification everyo/validation.py
Benchmarking everyo/benchmarking.py

Other components

Developer

Connections 29

Developer → Data Preparation: prepares data README.md
Data Preparation → Dataset Abstractions: builds datasets
Developer → Datasets: chooses data
Dataset Abstractions → Batch Loading: feeds batches everyo/data/dataset.py
Developer → Tensors and Autograd: creates tensors README.md
Tensors and Autograd → Differentiable Operations: uses operations README.md
Differentiable Operations → Gradient Engine: records gradients everyo/core/operations.py
Differentiable Operations → NumPy Backend: dispatches computation
Developer → Neural Network Layers: defines model README.md
Embedding Layer → Differentiable Operations: calls indexing everyo/nn/embedding.py
Neural Network Layers → Loss Functions: computes loss README.md
Training Loop → Batch Loading: reads batches everyo/training/trainer.py
Training Loop → Gradient Engine: uses gradient control everyo/training/trainer.py
Training Loop → Optimizers: updates parameters README.md
Training Loop → Neural Network Layers: trains model everyo/training/trainer.py
Training Loop → Evaluation: evaluates model
Developer → Evaluation: evaluates results README.md
Developer → Inference: requests predictions
Developer → Save and Load: saves or loads README.md
Developer → ONNX Export: exports model README.md
Developer → Dynamic Quantization: quantizes model README.md
Differentiable Operations → CUDA Operations: dispatches kernels everyo/core/operations.py
CUDA Operations → CUDA Bindings: calls bindings everyo/cuda/interface.py
Developer → TensorFlow Reference: cross-checks numerics README.md
Developer → Experiment Tools: runs experiments
Experiment Tools → Experiment Tracking: records runs
Developer → Monitoring: observes runtime
Developer → Validation and Verification: checks behavior README.md
Developer → Benchmarking: measures performance README.md

Drawn by GitDiagram on November 10, 2026 from the repository's file tree, README and source files. Read as Markdown ·