Instructions to use yunicro/MiDM-9B-q35-e1-bx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use yunicro/MiDM-9B-q35-e1-bx with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
File size: 1,367 Bytes
5e7e925 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 | # v0.2.2 — ScenarioView report and financial validation
Documentation/report release dated 2026-10-07. Model weights, tokenizer and loader are unchanged from v0.2.0; this is NOT an improved-accuracy model release.
- Publishes a mobile-friendly ScenarioView report using the previous local report layout styles and section order.
- Includes public macro observations with observation dates and source attribution, RT definitions, aggregate action metrics, and section-level provenance.
- Frozen MiDM action accuracy: 54.37%; B3+MiDM stacking: 50.93%; development-selected horizon/position residual: 47.49%. No candidate promoted.
- Evaluation: KOSPI, 2024-01-02 to 2026-08-20, 756 hypothetical decisions / 129 sampled dates, previously seen retrospective data; not independent trades or portfolio returns.
- No validated economic leading/coincident/lagging status. New prompt context packs are prepared locally but have not been inference-tested.
- No private source prices, per-date predictions, generated private narratives, checkpoint files or credentials published.
- The report is issued today but does not contain a freshly inferred today-market signal. No automatic daily schedule is configured.
Report: https://yeohoonyun.github.io/midm-decision-models/
The prior Zenodo DOI continues to refer to v0.2.1; no new Zenodo deposit is claimed.
|