flexnet.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# import json\n",
    "# from pprint import pprint\n",
    "\n",
    "# from torch.nn import Conv2d, MaxPool2d, AdaptiveAvgPool2d, ReLU, Sequential\n",
    "\n",
    "# from lib.fully_connected_net import FullyConnectedNet\n",
    "# from lib.flatten import Flatten\n",
    "\n",
    "\n",
    "# fname = 'DNNs/alexnet_2_65_120190602141824593525_created/k_4/model_params.json'\n",
    "\n",
    "# with open(fname) as json_file:\n",
    "#     layers = json.load(json_file)\n",
    "    \n",
    "# pprint(layers)\n",
    "\n",
    "# modules = []\n",
    "\n",
    "# for layer in layers:\n",
    "#     if layer['type'] == 'conv2d':\n",
    "#         module = Conv2d(layer['in_channels'],\n",
    "#                         layer['out_channels'],\n",
    "#                         (layer['kernel_height'], layer['kernel_width']),\n",
    "#                         stride=(layer['stride_height'], layer['stride_width'])\n",
    "#                        )\n",
    "#     if layer['type'] == 'maxpool2d':\n",
    "#         module = MaxPool2d((layer['kernel_height'], layer['kernel_width']),\n",
    "#                            stride=(layer['stride_height'], layer['stride_width']),\n",
    "#                           )\n",
    "    \n",
    "#     if layer['type'] == 'adaptiveavgpool2d':\n",
    "#         module = AdaptiveAvgPool2d((layer['out_height'], layer['out_width']))\n",
    "        \n",
    "#     if layer['type'] == 'fcs':\n",
    "#         flatten = Flatten()\n",
    "#         modules.append(flatten)\n",
    "#         fcs_dropout = 0\n",
    "#         batch_norm = True\n",
    "#         module = FullyConnectedNet(layer['input_size'], layer['output_size'], fcs_dropout, \\\n",
    "#                                    batch_norm, layer['width'], layer['num_layers'])\n",
    "        \n",
    "        \n",
    "#     modules.append(module)\n",
    "    \n",
    "#     if layer['type'] == 'conv2d':\n",
    "#         modules.append(ReLU(inplace=True))\n",
    "        \n",
    "    \n",
    "    \n",
    "# modules\n",
    "\n",
    "# Sequential(*modules)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'batch_size': 32,\n",
      " 'cost_function': 'L1',\n",
      " 'data_is_target': 0,\n",
      " 'data_noise_gaussian': 1,\n",
      " 'data_train': '/Users/zhanwenchen/Downloads/20180402_L74_70mm/train_1.h5',\n",
      " 'data_val': '/Users/zhanwenchen/Downloads/20180402_L74_70mm/val_1.h5',\n",
      " 'input_dims': [2, 65, 1],\n",
      " 'k': 4,\n",
      " 'layers': [{'in_channels': 1,\n",
      "             'kernel_height': 5,\n",
      "             'kernel_width': 66,\n",
      "             'out_channels': 194,\n",
      "             'padding_height': 2,\n",
      "             'padding_width': 3,\n",
      "             'stride_height': 4,\n",
      "             'stride_width': 5,\n",
      "             'type': 'conv2d'},\n",
      "            {'kernel_height': 1,\n",
      "             'kernel_width': 2,\n",
      "             'stride_height': 4,\n",
      "             'stride_width': 3,\n",
      "             'type': 'maxpool2d'},\n",
      "            {'in_channels': 194,\n",
      "             'kernel_height': 3,\n",
      "             'kernel_width': 1,\n",
      "             'out_channels': 96,\n",
      "             'padding_height': 1,\n",
      "             'padding_width': 1,\n",
      "             'stride_height': 2,\n",
      "             'stride_width': 1,\n",
      "             'type': 'conv2d'},\n",
      "            {'kernel_height': 1,\n",
      "             'kernel_width': 2,\n",
      "             'stride_height': 2,\n",
      "             'stride_width': 1,\n",
      "             'type': 'maxpool2d'},\n",
      "            {'in_channels': 96,\n",
      "             'kernel_height': 2,\n",
      "             'kernel_width': 4,\n",
      "             'out_channels': 157,\n",
      "             'padding_height': 1,\n",
      "             'padding_width': 3,\n",
      "             'stride_height': 1,\n",
      "             'stride_width': 3,\n",
      "             'type': 'conv2d'},\n",
      "            {'in_channels': 157,\n",
      "             'kernel_height': 1,\n",
      "             'kernel_width': 7,\n",
      "             'out_channels': 56,\n",
      "             'padding_height': 0,\n",
      "             'padding_width': 3,\n",
      "             'stride_height': 5,\n",
      "             'stride_width': 3,\n",
      "             'type': 'conv2d'},\n",
      "            {'in_channels': 56,\n",
      "             'kernel_height': 1,\n",
      "             'kernel_width': 3,\n",
      "             'out_channels': 186,\n",
      "             'padding_height': 0,\n",
      "             'padding_width': 2,\n",
      "             'stride_height': 1,\n",
      "             'stride_width': 2,\n",
      "             'type': 'conv2d'},\n",
      "            {'kernel_height': 1,\n",
      "             'kernel_width': 2,\n",
      "             'stride_height': 5,\n",
      "             'stride_width': 2,\n",
      "             'type': 'maxpool2d'},\n",
      "            {'out_height': 1, 'out_width': 1, 'type': 'adaptiveavgpool2d'},\n",
      "            {'input_size': 186,\n",
      "             'num_layers': 1,\n",
      "             'output_size': 130,\n",
      "             'type': 'fcs',\n",
      "             'width': 202}],\n",
      " 'learning_rate': 6.557821399239181e-05,\n",
      " 'optimizer': 'Adam',\n",
      " 'patience': 20,\n",
      " 'type': 'alexnet',\n",
      " 'version': '1.0',\n",
      " 'weight_decay': 0.022919212077948302}\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Sequential(\n",
       "  (0): PrintLayer()\n",
       "  (1): Conv2d(1, 194, kernel_size=(5, 66), stride=(4, 5))\n",
       "  (2): ReLU(inplace)\n",
       "  (3): PrintLayer()\n",
       "  (4): MaxPool2d(kernel_size=(1, 2), stride=(4, 3), padding=0, dilation=1, ceil_mode=False)\n",
       "  (5): PrintLayer()\n",
       "  (6): Conv2d(194, 96, kernel_size=(3, 1), stride=(2, 1))\n",
       "  (7): ReLU(inplace)\n",
       "  (8): PrintLayer()\n",
       "  (9): MaxPool2d(kernel_size=(1, 2), stride=(2, 1), padding=0, dilation=1, ceil_mode=False)\n",
       "  (10): PrintLayer()\n",
       "  (11): Conv2d(96, 157, kernel_size=(2, 4), stride=(1, 3))\n",
       "  (12): ReLU(inplace)\n",
       "  (13): PrintLayer()\n",
       "  (14): Conv2d(157, 56, kernel_size=(1, 7), stride=(5, 3))\n",
       "  (15): ReLU(inplace)\n",
       "  (16): PrintLayer()\n",
       "  (17): Conv2d(56, 186, kernel_size=(1, 3), stride=(1, 2))\n",
       "  (18): ReLU(inplace)\n",
       "  (19): PrintLayer()\n",
       "  (20): MaxPool2d(kernel_size=(1, 2), stride=(5, 2), padding=0, dilation=1, ceil_mode=False)\n",
       "  (21): PrintLayer()\n",
       "  (22): AdaptiveAvgPool2d(output_size=(1, 1))\n",
       "  (23): Flatten()\n",
       "  (24): PrintLayer()\n",
       "  (25): FullyConnectedNet(\n",
       "    (layers): ModuleList(\n",
       "      (0): Linear(in_features=186, out_features=202, bias=True)\n",
       "      (1): Linear(in_features=202, out_features=130, bias=True)\n",
       "    )\n",
       "    (batch_norm_layers): ModuleList(\n",
       "      (0): BatchNorm1d(202, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    )\n",
       "    (relu): ReLU()\n",
       "    (dropout): Dropout(p=0)\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from lib.flexnet import FlexNet\n",
    "\n",
    "fname = 'DNNs/alexnet_2_65_120190604164319588038_created/k_4/model_params.json'\n",
    "\n",
    "\n",
    "model = FlexNet(fname)\n",
    "model.net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}