xiaden/ML-Experiments

adaptive-token-budget

Methods for adaptive/token-budgeted visual generation — token merging, token pruning, early exiting, and self-budgeting tokenization in diffusion/image generation models.

attnres-implementation

Attention Residuals (AttnRes) implementation in the PDLI experiment. Covers AttnResDepth, AttnResBlock, and the L2RTP model that uses them. Essential for anyone modifying the transformer architecture or residual connections.

check-dependencies

Comprehensive dependency maintenance — verifies import health, requirements.txt consistency, import hierarchy, and environment health.

check-paper-library

Comprehensive paper library maintenance — cross-references papers.md against summaries, techniques, and PDFs. Detects orphans, missing registrations, and pulls missing papers.

add-research-paper

Add a new research paper to the library — fetch the paper, write a structured summary to `docs/upstream/papers/`, implement its technique as a standalone Python module in `resources/techniques/`, and register it in `references/papers.md`.

pyramidal-diffusion

Multi-scale/pyramidal approaches to denoising diffusion models. Covers the Ryu & Ye 2022 pyramidal DDPM paper and related coarse-to-fine diffusion architectures. Relevant to subdivision mode / spatial budget training and hierarchical latent backbones in the…

validate-artifacts

Comprehensive artifact maintenance — validates log naming, decision format, plan syntax compliance, staleness detection, and ensures the artifact corpus is parseable.

validate-datasets

Dataset management maintenance — validates dataset integrity, version tracking, storage health, and cache consistency for training datasets.

validate-gpu-health

GPU and VRAM health maintenance — validates CUDA availability, monitors VRAM usage, detects memory leaks, and ensures GPU readiness for training.

validate-mlflow

MLflow experiment tracking maintenance — validates run metadata, experiment organization, artifact consistency, and SQLite backend health.

validate-model-artifacts

Model checkpoint maintenance — validates .pt file integrity, metadata structure, version tracking, and storage health for trained disposition layers.

validate-notebooks

Validates standard/experimental notebook conventions, shared component usage, and lifecycle hooks. Provides grep-based checks and targeted auto-fix guidance.

validate-techniques

Technique module maintenance — validates conventions, checks paper pulls, detects duplicates, verifies registration, and auto-fixes import issues.

validate-training-pipeline

Training pipeline maintenance — validates experiment runner, training loop, training rig, and component integration for all notebooks.