bkm 0.0.0

BKM (Best Known Method) DAG generator for wafer process flows

bkm

BKM (Best Known Method) DAG generator for wafer process flows.

Build, validate, fingerprint, and match Directed Acyclic Graphs that describe semiconductor wafer process flows from parquet data.

Installation

pip install bkm
pip install bkm[viz]   # with graphviz rendering

Quick start (Python API)

from bkm import BKM

bkm = BKM(
    version="v1",
    process_data={
        "in": ["a"],
        "a":  ["b", "c"],
        "b":  ["out"],
        "c":  ["out"],
    },
)

print(bkm.fingerprint())
print(bkm.print_graph())
bkm.save("v1.json")

CLI usage (bkm-generator)

Process all parquet files in a folder and assign BKM versions:

bkm-generator --data_folder ./training_data --update

Subsequent runs assign wafers to existing BKMs and flag unmatched as unknown:

bkm-generator --data_folder ./new_data

Documentation

• bkm.md — BKM class API, DAG concepts, edge inference algorithm
• bkm_generator.md — multi-BKM registry, loaf composition, CLI pipeline

License

MIT © 2026 yRocket