Clayton Young
c7c5bc21ff
feat(example): add working TimesFM forecast example with global temperature data
...
- Add NOAA GISTEMP global temperature anomaly dataset (36 months, 2022-2024)
- Run TimesFM 1.0 PyTorch forecast for 2025 (12-month horizon)
- Generate fan chart visualization with 80%/90% confidence intervals
- Create comprehensive markdown report with findings and API notes
API Discovery:
- TimesFM 2.5 PyTorch checkpoint has file format issue (model.safetensors
vs expected torch_model.ckpt)
- Working API uses TimesFmHparams + TimesFmCheckpoint + TimesFm() constructor
- Documented API in GitHub README differs from actual pip package
Includes:
- temperature_anomaly.csv (input data)
- forecast_output.csv (point forecast + quantiles)
- forecast_output.json (machine-readable output)
- forecast_visualization.png (LFS-tracked)
- run_forecast.py (reusable script)
- visualize_forecast.py (chart generation)
- run_example.sh (one-click runner)
- README.md (full report with findings)
2026-02-23 07:43:04 -05:00
Clayton Young
98670bcf47
feat(skill): add timesfm-forecasting skill for time series forecasting
...
Add comprehensive TimesFM forecasting skill with mandatory system
preflight checks (RAM/GPU/disk), end-to-end CSV forecasting script,
full API reference, data preparation guide, and hardware requirements
documentation. Supports TimesFM 2.5 (200M), 2.0 (500M), and legacy
v1.0 with automatic batch size recommendations based on hardware.
2026-02-23 07:43:04 -05:00
