Cut the next Ariadne release so downstream consumers can adopt the configurable targeted k-CFA depth and the per-context tensor-shape precision it enables.
ponder-lab#371 (implementing #530 and #379) adds a configurable targeted k-CFA depth to PythonTensorAnalysisEngine and restores precise per-context layer-output shapes for the model-forward archetype. None of this is reachable by consumers until it ships in a release and they bump their Ariadne dependency. The current development version is 0.47.1-SNAPSHOT.
Cut the next Ariadne release so downstream consumers can adopt the configurable targeted k-CFA depth and the per-context tensor-shape precision it enables.
ponder-lab#371 (implementing #530 and #379) adds a configurable targeted k-CFA depth to
PythonTensorAnalysisEngineand restores precise per-context layer-output shapes for the model-forward archetype. None of this is reachable by consumers until it ships in a release and they bump their Ariadne dependency. The current development version is0.47.1-SNAPSHOT.