|
Download README.md from yyoraa/drcik-forecast-cache: direct link, hf CLI and curl.
- Browser
- Download file 1.05 kB
-
https://huggingface.co/datasets/yyoraa/drcik-forecast-cache/resolve/main/README.md
- Command line
-
hf download hf://datasets/yyoraa/drcik-forecast-cache/README.md
-
curl -L -o README.md https://huggingface.co/datasets/yyoraa/drcik-forecast-cache/resolve/main/README.md
1.05 kB
Dr-CiK forecast cache for the self-evolving time-series pipeline
champion_forecasts_gpt56sol_high_toto_balanced_v3_20260906.tar.gz (19 MB compressed, 199 MB / 49,340 JSON files unpacked) is the ForecastStore cache that scripts/nrd_precompute.py in
kaichen-z/time-series @ self-evolving-nrd reads with cache_only=True.
mkdir -p runs/champion_forecasts
tar -xzf champion_forecasts_gpt56sol_high_toto_balanced_v3_20260906.tar.gz -C runs/champion_forecasts
Each file is one cached forecast (fields forecast, status, key, metric_policy, …), keyed by method, input history, horizon, frequency and portfolio identity. The pipeline uses the future forecasts of the 31 methods listed in inputs/full_methods.json on the 199 Dr-CiK public tasks; the remaining entries come from other experiments. No evaluation labels are stored.
sha256 of the tarball: 30d2e1b56971682e1de1bd106e3028c35d67b00f60e1a5f520134e7d8f2adc46