The NPZ package provides support for reading and writing Numpy .npy and .npz files in Julia.

An .npy file contains a single numpy array, stored in a binary format along with its shape, data type, etc. An .npz file contains a collection numpy arrays each encoded in the .npy format and stored in a ZIP file. For more information, see the numpy.save, numpy.savez, numpy.savez_compressed, and numpy.load functions in the Numpy documentation.

Install via the Julia package manager, Pkg.add("NPZ").

We try to provide an interface similar to the MAT.jl package. Some usage examples are shown below.

Write and then read back an array:

julia> using NPZ

julia> x = [i-j for i in 1:3, j in 1:4];

julia> npzwrite("x.npy", x)

julia> y = npzread("x.npy")
3x4 Int64 Array:
 0  -1  -2  -3
 1   0  -1  -2
 2   1   0  -1

Write and then read back a collection of arrays or numbers:

julia> npzwrite("data.npz", Dict("x" => x, "a" => collect(1:9), "n" => 42))

julia> vars = npzread("data.npz")
Dict{String,Any} with 3 entries:
  "x" => [0 -1 -2 -3; 1 0 -1 -2; 2 1 0 -1]
  "a" => [1, 2, 3, 4, 5, 6, 7, 8, 9]
  "n" => 42