CompressedTensorModel

field

Type: str [Required]

Base-64-encoded data in litte endian format.

Bool arrays are bitpacked, other types are IEEE753. Everything is little-endian. Tensors are C-ordering.

Validated by

• check_sizes

field

Type: Literal['f16', 'f32', 'f64', 'bool', 'i8', 'i16', 'i32', 'i64', 'u8', 'u16', 'u32', 'u64', 'c64', 'c128'] [Required]

The data type of the tensor.

Validated by

• check_sizes

field

Type: list[int] [Required]

The shape of the tensor.

Validated by

• check_sizes

validator check_sizes

Cross-validate that all sizes are consistent.

classmethod from_numpy(array)

Instantiate from a NumPy array.

Parameters

array (ndarray)

to_numpy()

Convert to a NumPy Array.

Return type

ndarray