libpython-clj2.python.np-array

Bindings for deeper intergration of numpy into the tech.v3.datatype system. This allows seamless usage of numpy arrays in datatype and tensor functionality such as enabling the tech.v3.tensor/ensure-tensor call to work with numpy arrays -- using zero copying when possible.

All users need to do is call require this namespace; then as-jvm will convert a numpy array into a tech tensor in-place.

datatype->ptr-type-name

(datatype->ptr-type-name dtype)

descriptor->numpy

(descriptor->numpy {:keys [ptr shape strides elemwise-datatype], :as buffer-desc})

dtype->py-dtype-map

numpy->desc

(numpy->desc np-obj)

obj-dtype->dtype

(obj-dtype->dtype py-dtype)

py-dtype->dtype-map