Reference

The API, page by page, with every public name.

• Operations: the operations, grouped by category.

• Results: what operations return, and any_of.

• Data: arrays, exact values, formats and rounding.

• Algorithms: named implementations of operations.

• PyTorch equivalents: which ExactRef call checks which PyTorch call.

• Errors: the exceptions ExactRef raises.

Every public name

Name	Page
er.matmul, er.gemm	matmul, gemm
er.matrix_exp, er.matrix_exp_frechet	matrix_exp and matrix_exp_frechet
er.add, er.multiply	add and multiply
er.sum	sum
er.scaled_dot_product_attention	scaled_dot_product_attention
er.Result, er.Outputs, er.Boundary	Result
er.ExactArray, er.Enclosure	ExactArray and Enclosure
er.any_of	any_of
er.asarray, er.Array, er.MutableArray, er.FloatValue, er.ComplexValue	Arrays and values
er.FORMATS, er.get_format, er.Rounding, the format classes, and each format by name (er.float32, ...)	Formats and rounding
matmul_sequential, matmul_pairwise, matmul_blocked, matmul_exact_accumulation, matmul_slots16 (in er.linalg.products)	Algorithms
er.nn.attention.AttentionAlgorithm	Algorithms
er.Algorithm, er.ErrorModel, exactref.accumulation	Algorithms
er.ReferenceUnavailableError, er.UnsupportedFormatError	Errors

The operations are also under their category: er.linalg.products (matmul, gemm), er.linalg.matrix_function (matrix_exp, matrix_exp_frechet), er.arithmetic.basic (add, multiply), er.reduction.arithmetic (sum) and er.nn.attention (scaled_dot_product_attention).