Threading¶
Turn notebooks into high-performance research pipelines.
Threading transforms your notebooks and scripts into a reproducible, accelerated workspace—tracking provenance, enabling safe parameter sweeps, and optimizing execution.
$ threading init
[scan] Project layout detected
- notebooks/: 3 notebooks
- src/: preprocessing.py, pca.py, utils.py
[profile] Profiling notebooks & scripts
- 03_pca_analysis.ipynb
• Cell 8 (PCA): 42.1s
• Shape: ~30k samples × 8k features
[optimize] Targeted optimization
- Vectorizing PCA kernel
- GPU backend available (auto)
[provenance] Capturing experiment lineage
- Code hash: src/pca.py@a13f9c
- Input data: counts_filtered.tsv (sha256)
✔ Workspace ready
How it works¶
Step 1: Intelligent Discovery
Scans requirements.txt and notebook outputs to understand your experiment.
Step 2: Implicit Parallelism
Maps your logic to optimized GPU kernels without writing CUDA.
Step 3: Reproducible Scale
Run hardened workflows on terabytes with a single command.