Model Checkpoint Archives — Monumental Systems
Collection
Public preservation archives for raw checkpoints and lineage evidence; not supported inference releases. • 5 items • Updated
Harmonic GPT local checkpoint archive
Public preservation copy of selected local Harmonic GPT experiment artifacts
from /tank/harmonic-gpt-artifacts. This repository is an archival research
record, not a supported model release or a blanket grant to reuse every run.
Archive only. Public visibility preserves checkpoints and provenance. Use the separately packaged model repositories for supported inference and for lineage-specific licensing information.
Status
- Individual runs remain pending per-run provenance, data-license, and release-quality review.
- Raw training checkpoints may include optimizer state and machine-specific paths. Use the separately packaged public model repositories for inference.
- Evaluation-contaminated runs must remain clearly labeled and must not be presented as generalization results.
- Logs, PID files, caches, and other operational scratch files are excluded from the checkpoint upload.
Public release policy
Models promoted from this archive should receive their own Hugging Face model repository with:
- inference-ready weights and configuration;
- architecture and attention/state-mixing classification;
- training-data summary and license;
- evaluation methodology, including contamination disclosure;
- intended-use and limitation notes;
- an immutable source commit and reproducibility instructions.
The Monumental Systems website is the curated release index, Hugging Face is the weight/model-card distribution layer, and GitHub remains the source and reproduction layer.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support