Coverage

This page collects links to all coverage and supported-operator pages across the different conversion backends.

SQL / DataFrame / Polars

SQL / DataFrame Coverage summarises which SQL constructs, DataFrame operations, and polars LazyFrame operations are currently supported when converting tabular data manipulations to ONNX.

scikit-learn and compatible libraries

Supported Converters lists all scikit-learn estimators and transformers, along with estimators from category_encoders, imbalanced-learn, lightgbm, scikit-survival, statsmodels, and xgboost, showing which ones have a registered converter in yobx.sklearn.

TensorFlow / JAX

Supported TF Ops lists every TensorFlow op that has a converter to ONNX.

Supported JAX Ops lists the JAX / StableHLO ops that are handled during JAX-to-ONNX conversion.

LiteRT

LiteRT / TFLite Export to ONNX describes the overall LiteRT/TFLite to ONNX conversion workflow, including dynamic shapes and custom op converters.

Supported LiteRT Ops lists every LiteRT (TFLite) op that has a converter to ONNX.

PyTorch

Supported Aten Functions enumerates every ATen function and its mapping to an ONNX operator.

Torch Export to ONNX also contains an overview of exportability (Overview of Exportability Comparison) that runs a broad set of model cases through multiple exporters and reports which ones succeed.

Op-db Coverage per Op and Type shows which op_db ops and data types are covered by the op-db export tests, distinguishing between ops with a working converter, known failures, and ops missing a converter.