Type information

tensorplay.finfo and tensorplay.iinfo describe the limits of a floating-point or integer data type. You can query a dtype, or construct one from a tensor.

import tensorplay as tp

print(tp.finfo(tp.float32))          # machine limits for float32
print(tp.finfo(tp.float32).max)      # 3.4028235e+38
print(tp.iinfo(tp.int64).max)        # 9223372036854775807

finfo works for floating-point dtypes (float16, float32, float64, and the complex dtypes) and exposes values such as max, min, eps, tiny, and resolution. iinfo works for integer dtypes (int8 through int64, and the unsigned types) and exposes min and max.

Both accept a dtype or the data type of a tensor:

x = tp.tensor([1.0, 2.0])
print(tp.finfo(x.dtype).eps)         # machine epsilon for float32