QUANTUM WAVEFull training trace
01 / OVERVIEWWaiting for measured training samples
Overview preserves sampled extrema as the trace grows
QWAVE / COMPUTATIONAL RESEARCHEXPERIMENT 002
Watch the wavefunction learn
A live view into neural quantum state optimization
EXPERIMENT STATEAwaiting telemetryConnecting to the research feed
H₂ / CONTINUOUS SPACEFERMINET + VMCGPU WORKERSAMPLER INITIALIZING
OPTIMIZATION STEPS—Actual completed iterations
ELAPSED RUNTIME—Data uploaded and refreshed every 15 seconds
ROLLING ENERGY / Eₕ—Mean over 50 training steps
SAMPLER ACCEPTANCE—Metropolis move acceptance
Inside the optimization
Training energy50-step rolling mean
Recent 500 steps
02 / DETAILWaiting for measured training samples
Recent samples are shown at their original resolution
Training diagnostics from a changing model — not an independently evaluated energy or a confidence interval
TRACEABLE BY DESIGN
Every run leaves evidence
Real samples travel from the GPU worker to this page
Model snapshots preserve the work behind the curve
- MODEL SNAPSHOT
- Awaiting first checkpoint
- REMOTE BACKUP
- Awaiting confirmation
- MODEL CONFIGURATION
- Awaiting configuration
- SCIENTIFIC REVIEW
- Independent evaluation pending
- RUN IDENTIFIER
- —
- BACKUP SHA256
- —
How to read this experiment
The model represents a neutral hydrogen molecule at a fixed bond length of 1·4 bohr with nuclear repulsion included in the energy
The raw line shows sampled training energy while the brighter line averages the latest 50 steps — fluctuations can persist as the model and samples change
A lower individual sample does not establish a better ground-state estimate — fixed-model sampling with uncertainty analysis is a separate evaluation stage
The feed is checked every 15 seconds — if fresh worker observations stop arriving the page marks the interruption and preserves the last measured trace
Read the evaluation protocol ↗