You can deploy this model and start earning money today!

Byteroot/
lane-rate-lgbm

Name: lane-rate-lgbm
Availability: InStock

Views

No views yet

Lane Rate (LightGBM)

Frozen LightGBM regressor for posted truck-load rates (posted_rate, USD).

This is not a Transformers checkpoint. Inference runs on the Docker Space Byteroot/lane-rate, which loads rate_model.joblib and a copy of the freight feature builder.

Files

rate_model.joblib — dict with builder, booster, family, params, metrics
current.json — same metadata without the fitted objects

Training snapshot

Family: LightGBM, 409 trees, seed 42
Target: log1p(posted_rate), MAE objective, inverted with expm1
Fit on: all 48,000 labeled rows (January–October 2025)
Selected on: January–August train / September–October holdout
Holdout MAE: $108.58

Load

The pickle includes freight.features.FeatureBuilder. You need that package on PYTHONPATH (same module path as training). Do not reimplement transform().

python

from huggingface_hub import hf_hub_download
import joblib

path = hf_hub_download("Byteroot/lane-rate-lgbm", "rate_model.joblib")
artifact = joblib.load(path)  # keys: builder, booster, family, ...

Training code: github.com/ProsperAsiimwe/spotter

Do not use this artifact as a live market feed. December rows have no market_index or quote_signal; calendar features are what remain.