test/tutorials/VGG.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "using Knet;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "KnetOnnx_PATH = \"/Users/egeersu/Desktop/KnetOnnx.jl/src\"\n",
    "push!(LOAD_PATH, KnetOnnx_PATH)\n",
    "using KnetOnnx;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Test: vgg16\n",
    "file_path = \"/Users/egeersu/Desktop/KnetOnnx.jl/test/ONNX_files/vgg16.onnx\"\n",
    "graph = ONNXtoGraph(file_path)\n",
    "PrintGraph(graph)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = KnetModel(file_path);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1000×10 Array{Float64,2}:\n",
       " -1.31474   -0.861497   -1.29136   …  -1.25527    -0.748424  -1.00036 \n",
       " -0.810534  -0.311417   -0.37097      -0.218187   -0.660613  -0.611758\n",
       " -1.73626   -0.834154   -0.823934     -0.941861   -1.24095   -1.05418 \n",
       " -1.40207   -0.414046   -0.555744     -0.70076    -0.872822  -0.864427\n",
       " -0.388837   0.325223    0.177983      0.0590558   0.202324  -0.125493\n",
       " -0.674042  -0.0796126  -0.458447  …  -0.471942   -0.123057  -0.512125\n",
       " -0.157227   0.486135    0.266582      0.123495    0.410126  -0.021465\n",
       " -0.211488  -0.608784   -0.554852     -0.534473   -0.22968   -0.444548\n",
       " -0.215734  -0.439477   -0.589214     -0.644863   -0.157965  -0.286376\n",
       " -1.1073    -1.81002    -1.43138      -1.60634    -0.893495  -1.57079 \n",
       " -2.08202   -2.09914    -1.98064   …  -2.16407    -2.09235   -2.14884 \n",
       " -1.23889   -1.42806    -1.37892      -1.5432     -1.26689   -1.2836  \n",
       " -1.43186   -1.62997    -1.4946       -1.66496    -1.59659   -1.69574 \n",
       "  ⋮                                ⋱                                  \n",
       " -0.080518  -0.237006   -0.146481     -0.368172   -0.12195   -0.136324\n",
       " -0.140845  -0.648358   -0.418296     -0.396057   -0.260694  -0.389918\n",
       "  0.414012  -0.171491   -0.210932  …  -0.230226    0.653784   0.118375\n",
       " -0.335709  -0.494984   -0.688399     -0.443332   -0.389234  -0.648623\n",
       " -0.539058  -0.55482    -0.773885     -0.510106   -0.573949  -0.57762 \n",
       " -2.37917   -2.13492    -2.34252      -2.2504     -2.33533   -2.32732 \n",
       " -2.37501   -2.35411    -2.51472      -2.27652    -2.47493   -2.50537 \n",
       " -1.49953   -1.56427    -1.52076   …  -1.59372    -1.41644   -1.56358 \n",
       " -1.00025   -1.08201    -1.33952      -1.12226    -0.925758  -1.2551  \n",
       " -1.33477   -1.44967    -1.76683      -1.4636     -0.938723  -1.49696 \n",
       "  1.21056    0.899171    0.69233       0.864885    1.01064    0.859657\n",
       "  1.99104    2.08073     1.95311       1.94072     1.73825    1.90464 "
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x = randn(Float32,224,224,3,10)\n",
    "out = model(xi)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Dict{Any,Any} with 42 entries:\n",
       "  \"vgg0_pool0_fwd\"       => Float32[2.31927 0.0 … 0.0 1.00264; 0.0 0.0 … 2.0254…\n",
       "  \"vgg0_relu10_fwd\"      => Float32[0.0 0.0 … 0.0 0.509059; 0.0 0.0 … 0.0 0.0; …\n",
       "  \"vgg0_conv12_fwd\"      => Float32[0.0 0.0 … 0.0 0.0; 0.0 0.0 … 0.0 0.0; … ; 0…\n",
       "  \"data\"                 => Float32[-0.465692 0.128407 … 0.70165 -0.641882; 1.6…\n",
       "  \"vgg0_conv9_fwd\"       => Float32[0.0 0.0 … 0.0 0.0; 0.0 0.0 … 0.0 0.0; … ; 0…\n",
       "  \"vgg0_dense1_fwd\"      => [0.23642 0.0600647 … 0.182999 0.356688; -0.540888 -…\n",
       "  \"vgg0_relu5_fwd\"       => Float32[0.0 2.81849 … 8.09113 10.2458; 0.0 2.48117 …\n",
       "  \"vgg0_conv5_fwd\"       => Float32[0.0 2.81849 … 8.09113 10.2458; 0.0 2.48117 …\n",
       "  \"vgg0_relu12_fwd\"      => Float32[0.0 0.0 … 0.0 0.0; 0.0 0.0 … 0.0 0.0; … ; 0…\n",
       "  \"vgg0_relu4_fwd\"       => Float32[17.8061 16.9484 … 16.5784 15.3698; 6.27331 …\n",
       "  \"vgg0_dense0_relu_fwd\" => [0.32355 0.258957 … 0.304839 0.263062; 0.437306 0.3…\n",
       "  \"vgg0_conv2_fwd\"       => Float32[2.3094 0.0 … 1.88638 0.0; 2.01959 3.93871 ……\n",
       "  \"vgg0_relu1_fwd\"       => Float32[2.31927 1.32234 … 0.0 1.00264; 1.64323 0.34…\n",
       "  \"vgg0_conv3_fwd\"       => Float32[0.0 0.0 … 0.0 0.0; 0.149799 0.0 … 0.0 0.0; …\n",
       "  \"vgg0_conv6_fwd\"       => Float32[0.0 0.0 … 0.0 0.0; 0.0 3.44817 … 0.0 0.0; ……\n",
       "  \"vgg0_conv10_fwd\"      => Float32[0.0 0.0 … 0.0 0.509059; 0.0 0.0 … 0.0 0.0; …\n",
       "  \"vgg0_pool2_fwd\"       => Float32[3.44817 0.0 … 0.0 0.0; 0.0 0.0 … 0.0 0.0; ……\n",
       "  \"vgg0_pool1_fwd\"       => Float32[0.149799 10.165 … 0.0 0.0; 0.0 0.0 … 0.0 0.…\n",
       "  \"vgg0_relu11_fwd\"      => Float32[0.0 0.0 … 0.0 0.0; 0.0 0.0 … 0.0 0.0; … ; 0…\n",
       "  \"flatten_60\"           => Float32[0.0 0.0 … 0.0 0.0; 0.0 0.0 … 0.0 0.0; … ; 0…\n",
       "  \"vgg0_conv11_fwd\"      => Float32[0.0 0.0 … 0.0 0.0; 0.0 0.0 … 0.0 0.0; … ; 0…\n",
       "  \"vgg0_relu8_fwd\"       => Float32[5.20256 0.0 … 2.25456 0.0; 0.0 0.0 … 3.0918…\n",
       "  \"vgg0_dense0_fwd\"      => [0.32355 0.258957 … 0.304839 0.263062; 0.437306 0.3…\n",
       "  \"vgg0_conv7_fwd\"       => Float32[0.0 0.0 … 6.43491 0.0; 0.0 0.0 … 0.0 0.0; ……\n",
       "  \"vgg0_relu2_fwd\"       => Float32[2.3094 0.0 … 1.88638 0.0; 2.01959 3.93871 ……\n",
       "  ⋮                      => ⋮"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.tensors"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Julia 1.2.0",
   "language": "julia",
   "name": "julia-1.2"
  },
  "language_info": {
   "file_extension": ".jl",
   "mimetype": "application/julia",
   "name": "julia",
   "version": "1.2.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}