hamilton-v2

whitepaper and dataset

https://zenodo.org/records/20833711?token=eyJhbGciOiJIUzUxMiJ9.eyJpZCI6ImMyYmJiYjhmLTk0OTktNDk2Yi1hOGEzLTIwMGVmNjNhZDgyMSIsImRhdGEiOnt9LCJyYW5kb20iOiI4ODI4NTdiOTRmMjMwMGEyNjdmYzc5OWY5ZmQ2MGFlNCJ9.pMGUp40H_2YKYcm7Y7REHQBqw_LXiznPTGiUAk-pYkxWsa2TSClK3qPFMQaCBrE5QhImLmzZOxRkZHOkNQgSAw

https://zenodo.org/records/20832722?preview=1&token=eyJhbGciOiJIUzUxMiJ9.eyJpZCI6ImZhMWNlZjNjLWM2MzEtNGI4Ny04MmJhLWQ3MDA1YzE5YmVkNiIsImRhdGEiOnt9LCJyYW5kb20iOiI4MDYzZWYzZWQ0NTNhYTE4ZWNjNTVlOTg3ZDYwNGU4ZiJ9.a0C1HqBk_AN3mEQZQ_JhjbGlABkELMmNsj-kWFZXUu0tYVGeoRYQ36h_J9jQW4eyCZJxPby6WNR8UVOUaRWYJg

Real-Time Quantum Transformer Framework

This repository hosts the official architecture implementation and weights for the Real-Time Quantum Transformer Framework for Multidimensional Interaction Tracking and Emergent Phenomenon Prediction.

The model unifies multi-dimensional quantum tracking, structural datasets, and theoretical models into a single framework, bounding operational fields using geometric manifolds like organic molecular crystal trajectories (OMC25) and temporal discretization baselines (QM40).

Model Description

Instead of traditional dot-product self-attention mechanisms, this model implements Quantum Self-Attention acting over an isomorphic Hilbert space mapping.

• State Vector Mapping: Maps deep-feature representations via an MLP projection onto continuous latent state trajectories.
• Quantum Attention Overlaps: Calculates attention transition scores dynamically using state amplitude interference metrics: $| \langle Q_i | K_j \rangle |^2$.
• Downstream Tracking: Tracks non-linear, non-Euclidean quantum trajectories (e.g., Quantum 2 Curves) for forecasting structural deformations, localized stress, and multi-photon spectral dynamics.

Quickstart & Usage

To instantiate this model dynamically, make sure you have torch and transformers installed:

from transformers import AutoModel
import torch

# Load the model directly from the Hugging Face Hub
model = AutoModel.from_pretrained("GlimmaryKarl/quantum-transformer-v1", trust_remote_code=True)

# Example: Run a forward pass using a sequence of 12-dimensional tracking features
# Tensor shape: [Batch_size=1, Sequence_length=5, Input_dim=12]
sample_input = torch.randn(1, 5, 12)
output_logits = model(sample_input)

print("Output matrix shape:", output_logits.shape)