YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Graph-Residual with AMPLIFY-120M
This repository contains the AMPLIFY-120M Graph-Residual implementation and the official AMPLIFY-120M backbone bundle.
Contents
models/amplify/: official AMPLIFY-120M weights, configuration, tokenizer, and source files.src/models/amplify_graph_residual.py: Graph-Residual adaptation path.src/: model, graph, training, and evaluation code.configs/: portable AMPLIFY and evaluation configurations.
The verified P3 path projects 64-dimensional graph features to the 640-wide AMPLIFY representation, applies normalized gated residual fusion, and preserves the input sequence norm. Raw datasets, graph databases, logs, and optimizer state are not included.
Intended use
Protein representation adaptation, relation-aware retrieval, and analysis of heterogeneous biological graph context.
Limitations
Compatible graph representation inputs are required. This release is not a universal protein predictor and is not a generic unseen-node inductive model.
Citation
Please cite the associated Graph-Residual manuscript or preprint, the upstream AMPLIFY-120M model, and any upstream components included in the bundle.