yobx.tensorflow.ops.reduction_ops#

Converters for TF reduction and cumulative ops.

Reduction#

Sum (reduce_sum), Max (reduce_max), Min (reduce_min), Mean (reduce_mean), Prod (reduce_prod)

Variance / standard deviation#

tf.math.reduce_variance and tf.math.reduce_std are Python-level TensorFlow functions that decompose into primitive graph ops (Mean, SquaredDifference, Sqrt, etc.). Because every component op has a registered converter, these high-level functions are automatically supported without a dedicated single-op converter.

Cumulative#

Cumsum

yobx.tensorflow.ops.reduction_ops.convert_cumsum(g: GraphBuilderExtendedProtocol, sts: Dict[str, Any], outputs: List[str], op: Operation) → str[source]#

TF Cumsum (tf.math.cumsum) → ONNX CumSum.

yobx.tensorflow.ops.reduction_ops.convert_reduce_max(g: GraphBuilderExtendedProtocol, sts: Dict[str, Any], outputs: List[str], op: Operation) → str[source]#

TF Max (tf.math.reduce_max) → ONNX ReduceMax.

yobx.tensorflow.ops.reduction_ops.convert_reduce_mean(g: GraphBuilderExtendedProtocol, sts: Dict[str, Any], outputs: List[str], op: Operation) → str[source]#

TF Mean (tf.math.reduce_mean) → ONNX ReduceMean.

yobx.tensorflow.ops.reduction_ops.convert_reduce_min(g: GraphBuilderExtendedProtocol, sts: Dict[str, Any], outputs: List[str], op: Operation) → str[source]#

TF Min (tf.math.reduce_min) → ONNX ReduceMin.

yobx.tensorflow.ops.reduction_ops.convert_reduce_prod(g: GraphBuilderExtendedProtocol, sts: Dict[str, Any], outputs: List[str], op: Operation) → str[source]#

TF Prod (tf.math.reduce_prod) → ONNX ReduceProd.

yobx.tensorflow.ops.reduction_ops.convert_reduce_sum(g: GraphBuilderExtendedProtocol, sts: Dict[str, Any], outputs: List[str], op: Operation) → str[source]#

TF Sum (tf.math.reduce_sum) → ONNX ReduceSum.