dtype¶

class dragon.vm.torch.dtype(s)[source]¶

The basic data type.

Following data types are defined:

• torch.float16 or torch.half: 16-bit half-precision floating-point.
• torch.float32 or torch.float: 32-bit single-precision floating-point.
• torch.float64 or torch.double: 64-bit double-precision floating-point.
• torch.bfloat16: 16-bit truncated floating-point.
• torch.complex32: 32-bit single-precision complex.
• torch.complex64: 64-bit single-precision complex.
• torch.complex128: 128-bit double-precision complex.
• torch.int8: 8-bit signed integer.
• torch.uint8: 8-bit unsigned integer.
• torch.int16 or torch.short: 16-bit signed integer.
• torch.int32 or torch.int: 32-bit signed integer.
• torch.int64 or torch.long: 64-bit signed integer.
• torch.bool: Boolean.
• torch.qint8: Quantized 8-bit signed integer.
• torch.quint8: Quantized 8-bit unsigned integer.
• torch.qint32: Quantized 32-bit signed integer.