CIFAR10 with full dataset

cifar10_full.ml

#!/usr/bin/env owl

open Owl
open Owl_neural
open Algodiff.S
open Owl_neural_neuron

let model () = 
  let open Owl_neural_graph in
  let nn = input [|28;28;3|]
    |> conv2d [|3;3;3;32|] [|1;1|] ~act_typ:Activation.Relu
    |> conv2d [|3;3;32;32|] [|1;1|] ~act_typ:Activation.Relu ~padding:Owl_dense_ndarray_generic.VALID
    |> max_pool2d [|2;2|] [|2;2|] ~padding:Owl_dense_ndarray_generic.VALID
    |> dropout 0.25
    |> conv2d [|3;3;32;64|] [|1;1|] ~act_typ:Activation.Relu
    |> conv2d [|3;3;64;64|] [|1;1|] ~act_typ:Activation.Relu ~padding:Owl_dense_ndarray_generic.VALID
    |> max_pool2d [|2;2|] [|2;2|] ~padding:Owl_dense_ndarray_generic.VALID
    |> dropout 0.25
    |> fully_connected 1024 ~act_typ:Activation.Relu
    |> linear 10 ~act_typ:Activation.Softmax
    |> get_network
  in print nn;
  nn


let prepare_training_data () = 

  let datasets = Array.make 5 (Dense.Matrix.S.zeros 1 1) in 
  let labels   = Array.make 5 (Dense.Matrix.S.zeros 1 1) in 
  for i = 1 to 5 do 
      let x, y = Dataset.load_cifar_train_data i in 
      Array.set datasets (i-1) x;
      Array.set labels (i-1) y;
  done;

  let x = Dense.Matrix.S.concatenate datasets in 
  let y = Dense.Matrix.S.concatenate labels in 
  let m = Dense.Matrix.S.row_num x in 
  let x = Dense.Matrix.S.to_ndarray x in 
  let x = Dense.Ndarray.S.reshape x [|m;32;32;3|] in 
  (* let x = Dense.Ndarray.S.slice [[];[];[];[0]] x in *)
  let classes = 10 in 
  let y' = Dense.Matrix.S.zeros m classes in 
  for i = 0 to m - 1 do 
    Dense.Matrix.S.set y' i (int_of_float y.{i,0}) 1.
  done;
  let [|s1;s2;s3;s4|] = Dense.Ndarray.S.shape x in 
  let wy, hy = Dense.Matrix.S.shape y' in 
  Printf.printf "data shape: (%d, %d, %d, %d)\nlabels shape: (%d, %d).\n" s1 s2 s3 s4 wy hy;
  Dense.Ndarray.S.save x "cifar10_data";
  Dense.Matrix.S.save y' "cifar10_labels";
  ()

let train_cifar10_keras_graph () =

  let nn = model () in 
  (* let x, y = prepare_training_data () in *)
  let x = Dense.Ndarray.S.load "cifar10_data" in 
  let y = Dense.Matrix.S.load "cifar10_labels" in
  let params = Params.config
    ~batch:(Batch.Mini 32) ~learning_rate:(Learning_Rate.RMSprop (0.0001, 0.9)) ~checkpoint:0.1 1. in
  Graph.train_cnn ~params nn x y
  
let _ = 
  (* prepare_training_data () *)
  train_cifar10_keras_graph ()