examples/nitro_cnn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "2602fea3e061cd52",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:02.082535Z",
     "start_time": "2024-05-14T16:09:02.079798Z"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from tqdm.auto import trange\n",
    "import numpy as np\n",
    "import cupy as cp\n",
    "import pathlib\n",
    "import sys"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "f4be561f",
   "metadata": {},
   "outputs": [],
   "source": [
    "sys.path.append(str(pathlib.Path('../').resolve()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "845b705d213baff5",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:02.148660Z",
     "start_time": "2024-05-14T16:09:02.145430Z"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from lib.utils.data import Dataset, shuffle_dataset, load_integer_dataset\n",
    "from lib.utils.enums import Initialization, NonLinearity, OptimizerName\n",
    "from lib.utils.nn import accuracy, plot_history, l2_loss\n",
    "from lib.experiments import build_augmentation_pipeline\n",
    "from lib.utils.train import train_step_local, val_step\n",
    "from lib.utils.misc import to_cpu"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "14eaa1a8e3dcb853",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# Experiment config"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "13c93b79223a3377",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:02.153979Z",
     "start_time": "2024-05-14T16:09:02.149766Z"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "architecture = [\n",
    "    ('Conv2d', 64, 11, 1, 2),\n",
    "    ('MaxPool2d', 2, 2),\n",
    "    ('Conv2d', 192, 5, 1, 2),\n",
    "    ('MaxPool2d', 2, 2),\n",
    "    ('Conv2d', 384, 3, 1, 1),\n",
    "    ('Conv2d', 256, 3, 1, 1),\n",
    "    ('Conv2d', 256, 3, 1, 1),\n",
    "    ('MaxPool2d', 2, 2)\n",
    "]\n",
    "num_fc_layers = 2\n",
    "num_fc_hidden = [1024]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "ca06c48454741eb4",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:02.158061Z",
     "start_time": "2024-05-14T16:09:02.154988Z"
    }
   },
   "outputs": [],
   "source": [
    "dataset_config = dict(\n",
    "    dataset=Dataset.CIFAR10.name,\n",
    "    subsample_ratio=0.5,\n",
    "    ohe_values=(0, 32),\n",
    "    add_padding=False,\n",
    "    normalize=True,\n",
    "    data_augmentation=True,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "5f1bbb57c5407da5",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:02.167960Z",
     "start_time": "2024-05-14T16:09:02.164423Z"
    }
   },
   "outputs": [],
   "source": [
    "network_config = dict(\n",
    "    num_fc_layers=num_fc_layers,\n",
    "    num_fc_hidden=num_fc_hidden,\n",
    "    non_linearity=NonLinearity.NITRO_LEAKY_RELU.name,\n",
    "    init=Initialization.UNIFORM_KAIMING_LEAKY_RELU.name,\n",
    "    dtype='int32',\n",
    "    device='cuda',\n",
    "    local_loss='pred',\n",
    "    pred_decoder_dim=4096,\n",
    "    fe_dropout_rate=0.0,\n",
    "    fc_dropout_rate=0.25,\n",
    "    seed=42,\n",
    "    bias=False,\n",
    "    debug=False,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "fc63d5fe12448629",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:02.178686Z",
     "start_time": "2024-05-14T16:09:02.175531Z"
    }
   },
   "outputs": [],
   "source": [
    "training_config = dict(\n",
    "    batch_size=64,\n",
    "    num_epochs=35,\n",
    "    lr_inv=512,\n",
    "    lr_amp_factor=None,\n",
    "    fwd_decay_inv=15_000,\n",
    "    subnet_decay_inv=1_000,\n",
    "    optimizer=OptimizerName.INTEGER_SGD.name,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "76502789a59f337e",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:02.190658Z",
     "start_time": "2024-05-14T16:09:02.187692Z"
    }
   },
   "outputs": [],
   "source": [
    "config = dataset_config | network_config | training_config"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "f139d4d3466818c",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:02.213613Z",
     "start_time": "2024-05-14T16:09:02.209779Z"
    }
   },
   "outputs": [],
   "source": [
    "# Setup reproducibility\n",
    "np.random.seed(config[\"seed\"])\n",
    "cp.random.seed(config[\"seed\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7f13148687eb9acf",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "61c96d25da550e77",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:06.332713Z",
     "start_time": "2024-05-14T16:09:02.228153Z"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files already downloaded and verified\n",
      "Files already downloaded and verified\n",
      "Subsampling applied: training on 20000 samples\n",
      "Data normalized. New statistics:\n",
      "-  Min: [-123 -123 -105]\n",
      "-  Max: [126 127 127]\n",
      "- Mean: [0 0 0]\n",
      "-  Std: [61 62 61]\n",
      "\n",
      "Moved data to device: <CUDA Device 0>\n",
      "Train set:  (20000, 3, 32, 32), (20000, 10)\n",
      "Val set:    (10000, 3, 32, 32), (10000, 10)\n",
      "Test set:   (10000, 3, 32, 32), (10000, 10)\n",
      "Data types: (int8, int8)\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x2000 with 6 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "X_train, X_val, X_test, y_train, y_val, y_test = load_integer_dataset(config, add_padding=False, ohe_values=(0, 32))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "9988bf8f6add7a75",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:06.337319Z",
     "start_time": "2024-05-14T16:09:06.333718Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AugmentationPipeline([\n",
      "    RandomCrop(size=32, padding=4, probability=0.5)\n",
      "    RandomHorizontalFlip(probability=0.5)\n",
      "])\n"
     ]
    }
   ],
   "source": [
    "# Configure data augmentation\n",
    "augmentation_pipeline = build_augmentation_pipeline(config)\n",
    "print(augmentation_pipeline)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dd9d77152daf54c8",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "5d63f02301c65005",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:06.344788Z",
     "start_time": "2024-05-14T16:09:06.338201Z"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from lib.models.local_loss_models import IntegerLocalLossCNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "e4ce8a2a65ebaebe",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:06.358316Z",
     "start_time": "2024-05-14T16:09:06.345793Z"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "IntegerLocalLossCNN(\n",
      "\t(0): Sequential(\n",
      "\t\t\t(0): LocalConv2dBlock(\n",
      "\t\t\t(0): IntegerConv2d(in_channels=3, out_channels=64, bias=False, decay_rate_inv=15000)\n",
      "\t\t\t(1): ConstantScaling(fwd_factor=92928, bwd_factor=1)\n",
      "\t\t\t(2): NitroLeakyReLU(negative_slope_inv=4)\n",
      "\t\t\t(learning_layers): Sequential(\n",
      "\t\t\t(0): MaxPool2d(kernel_size=(4, 4), stride=(4, 4), padding=(2, 2))\n",
      "\t\t\t(1): Flatten()\n",
      "\t\t\t(2): IntegerLinear(in_features=3136, out_features=10, bias=False, decay_rate_inv=1000)\n",
      "\t\t\t(3): ConstantScaling(fwd_factor=802816, bwd_factor=1)\n",
      "\t\t)\n",
      "\t)\n",
      "\t\t\t(1): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=(0, 0))\n",
      "\t\t\t(2): LocalConv2dBlock(\n",
      "\t\t\t(0): IntegerConv2d(in_channels=64, out_channels=192, bias=False, decay_rate_inv=15000)\n",
      "\t\t\t(1): ConstantScaling(fwd_factor=409600, bwd_factor=1)\n",
      "\t\t\t(2): NitroLeakyReLU(negative_slope_inv=4)\n",
      "\t\t\t(learning_layers): Sequential(\n",
      "\t\t\t(0): MaxPool2d(kernel_size=(4, 4), stride=(4, 4), padding=(2, 2))\n",
      "\t\t\t(1): Flatten()\n",
      "\t\t\t(2): IntegerLinear(in_features=3072, out_features=10, bias=False, decay_rate_inv=1000)\n",
      "\t\t\t(3): ConstantScaling(fwd_factor=786432, bwd_factor=1)\n",
      "\t\t)\n",
      "\t)\n",
      "\t\t\t(3): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=(0, 0))\n",
      "\t\t\t(4): LocalConv2dBlock(\n",
      "\t\t\t(0): IntegerConv2d(in_channels=192, out_channels=384, bias=False, decay_rate_inv=15000)\n",
      "\t\t\t(1): ConstantScaling(fwd_factor=442368, bwd_factor=1)\n",
      "\t\t\t(2): NitroLeakyReLU(negative_slope_inv=4)\n",
      "\t\t\t(learning_layers): Sequential(\n",
      "\t\t\t(0): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=(0, 0))\n",
      "\t\t\t(1): Flatten()\n",
      "\t\t\t(2): IntegerLinear(in_features=3456, out_features=10, bias=False, decay_rate_inv=1000)\n",
      "\t\t\t(3): ConstantScaling(fwd_factor=884736, bwd_factor=1)\n",
      "\t\t)\n",
      "\t)\n",
      "\t\t\t(5): LocalConv2dBlock(\n",
      "\t\t\t(0): IntegerConv2d(in_channels=384, out_channels=256, bias=False, decay_rate_inv=15000)\n",
      "\t\t\t(1): ConstantScaling(fwd_factor=884736, bwd_factor=1)\n",
      "\t\t\t(2): NitroLeakyReLU(negative_slope_inv=4)\n",
      "\t\t\t(learning_layers): Sequential(\n",
      "\t\t\t(0): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=(0, 0))\n",
      "\t\t\t(1): Flatten()\n",
      "\t\t\t(2): IntegerLinear(in_features=2304, out_features=10, bias=False, decay_rate_inv=1000)\n",
      "\t\t\t(3): ConstantScaling(fwd_factor=589824, bwd_factor=1)\n",
      "\t\t)\n",
      "\t)\n",
      "\t\t\t(6): LocalConv2dBlock(\n",
      "\t\t\t(0): IntegerConv2d(in_channels=256, out_channels=256, bias=False, decay_rate_inv=15000)\n",
      "\t\t\t(1): ConstantScaling(fwd_factor=589824, bwd_factor=1)\n",
      "\t\t\t(2): NitroLeakyReLU(negative_slope_inv=4)\n",
      "\t\t\t(learning_layers): Sequential(\n",
      "\t\t\t(0): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=(0, 0))\n",
      "\t\t\t(1): Flatten()\n",
      "\t\t\t(2): IntegerLinear(in_features=2304, out_features=10, bias=False, decay_rate_inv=1000)\n",
      "\t\t\t(3): ConstantScaling(fwd_factor=589824, bwd_factor=1)\n",
      "\t\t)\n",
      "\t)\n",
      "\t\t\t(7): Dropout(dropout_rate=0.25)\n",
      "\t\t\t(8): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=(0, 0))\n",
      "\t\t)\n",
      "\t(1): IntegerLocalLossMLP(\n",
      "\t(0): Sequential(\n",
      "\t\t\t(0): Flatten()\n",
      "\t\t\t(1): LocalLinearBlock(\n",
      "\t\t\t(0): IntegerLinear(in_features=2304, out_features=1024, bias=False, decay_rate_inv=15000)\n",
      "\t\t\t(1): ConstantScaling(fwd_factor=589824, bwd_factor=1)\n",
      "\t\t\t(2): NitroLeakyReLU(negative_slope_inv=4)\n",
      "\t\t\t(learning_layers): Sequential(\n",
      "\t\t\t(0): IntegerLinear(in_features=1024, out_features=10, bias=False, decay_rate_inv=1000)\n",
      "\t\t\t(1): ConstantScaling(fwd_factor=262144, bwd_factor=1)\n",
      "\t\t)\n",
      "\t)\n",
      "\t\t\t(2): Dropout(dropout_rate=0.25)\n",
      "\t\t\t(3): Sequential(\n",
      "\t\t\t(0): IntegerLinear(in_features=1024, out_features=10, bias=False, decay_rate_inv=1000)\n",
      "\t\t\t(1): ConstantScaling(fwd_factor=262144, bwd_factor=1)\n",
      "\t\t)\n",
      "\t\t)\n",
      "\t)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "model = IntegerLocalLossCNN(\n",
    "    fe_architecture=architecture,\n",
    "    input_channels=X_train.shape[1],\n",
    "    image_size=X_train.shape[2],\n",
    "    num_fc_layers=config['num_fc_layers'],\n",
    "    num_hidden_fc=config['num_fc_hidden'],\n",
    "    num_classes=y_train.shape[1],\n",
    "    local_loss=config['local_loss'],\n",
    "    fe_dropout_rate=config['fe_dropout_rate'],\n",
    "    fc_dropout_rate=config['fc_dropout_rate'],\n",
    "    pred_decoder_dim=config['pred_decoder_dim'],\n",
    "    fwd_decay_inv=config['fwd_decay_inv'],\n",
    "    subnet_decay_inv=config['subnet_decay_inv'],\n",
    "    device=config['device'],\n",
    "    non_linearity=NonLinearity[config['non_linearity']],\n",
    "    init=Initialization[config['init']],\n",
    "    optimizer=OptimizerName[config['optimizer']],\n",
    "    dtype=config['dtype'],\n",
    "    bias=config['bias'],\n",
    "    debug=config['debug'],\n",
    "    lr_amp_factor=config['lr_amp_factor']\n",
    ")\n",
    "print(model.extra_repr(1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "17f45f7b0008d64e",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:06.374448Z",
     "start_time": "2024-05-14T16:09:06.359321Z"
    },
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Layer: conv_encoder_00\n",
      "- Weights:\n",
      "[[[[  1  -3   5 ...   2  -4   8]\n",
      "   [-10  10 -11 ...  13  -2   7]\n",
      "   [  9  -7   5 ...  -5 -13  -7]\n",
      "   ...\n",
      "   [  8  16 -16 ...  -3   4   4]\n",
      "   [  0 -13   6 ...   4  -6 -16]\n",
      "   [ -2 -16 -14 ... -16  11   5]]\n",
      "\n",
      "  [[  4  13  10 ... -10  -5 -15]\n",
      "   [ 15  14   2 ... -13 -16  15]\n",
      "   [ -8   8  10 ...  15 -12 -16]\n",
      "   ...\n",
      "   [ 11  11   8 ...   1   9  -8]\n",
      "   [ 15  -2  15 ...   5  15  15]\n",
      "   [-12  -8  -9 ...  -3  -7  -2]]\n",
      "\n",
      "  [[ -3 -14  -8 ...  12 -11   7]\n",
      "   [-11   7   0 ...   0  -5   9]\n",
      "   [ 14  -7  -3 ...  15  -9 -13]\n",
      "   ...\n",
      "   [  2  16  -8 ...   8   0   8]\n",
      "   [ 10   1 -12 ...  -4   4  10]\n",
      "   [ 11  -2 -10 ...   4  -6   2]]]\n",
      "\n",
      "\n",
      " [[[  3   4  -2 ...   8  12   9]\n",
      "   [  8 -14   2 ...  -6  12  -3]\n",
      "   [-10   2   6 ... -12  15   4]\n",
      "   ...\n",
      "   [  9  -9  11 ...  -1  -5 -16]\n",
      "   [-16   2   6 ...  13  -5  10]\n",
      "   [ -8 -15  -5 ...  -8  -3 -13]]\n",
      "\n",
      "  [[ 14   8  16 ...  -3  14   8]\n",
      "   [  9  -4 -14 ...   6   0  15]\n",
      "   [ -4   3 -11 ...  14  14 -15]\n",
      "   ...\n",
      "   [  9 -16  16 ...  -7   3   9]\n",
      "   [ 11  11  11 ...  14  13  11]\n",
      "   [  9   1   4 ...   8   7 -11]]\n",
      "\n",
      "  [[ -1  -8  16 ...   1  -5  -5]\n",
      "   [ 16  14  10 ...   5  -4  -7]\n",
      "   [ -5  14 -10 ...  -1   8  -1]\n",
      "   ...\n",
      "   [  4 -10  -6 ...  -4   8  -1]\n",
      "   [ 12  -5 -12 ...  -1   9  15]\n",
      "   [  9   7   5 ...   2  -3  12]]]\n",
      "\n",
      "\n",
      " [[[  3   3   2 ... -12   2 -14]\n",
      "   [ -3  -1   7 ... -16  -2  11]\n",
      "   [ -3   3  14 ...   5   9   5]\n",
      "   ...\n",
      "   [  1   1  15 ...   4 -13   1]\n",
      "   [-15  15   7 ... -14  -8   5]\n",
      "   [ -7 -14  -4 ...   9   6   8]]\n",
      "\n",
      "  [[  3   7  12 ... -13   7 -14]\n",
      "   [ -3   5  -3 ...   6  -8  13]\n",
      "   [ -1  -1 -10 ...   9   4   3]\n",
      "   ...\n",
      "   [  2   2 -14 ...  -7   6  -2]\n",
      "   [-14  15  -9 ...   2  -2 -16]\n",
      "   [ -7  -7  -3 ...   3   4   2]]\n",
      "\n",
      "  [[  5  13   1 ...  11  -3  13]\n",
      "   [-14  11  -4 ... -14   0 -13]\n",
      "   [  2 -13  11 ...   7  11  -8]\n",
      "   ...\n",
      "   [-13   5   9 ...  15 -16   8]\n",
      "   [-16   9  -6 ... -15   5  16]\n",
      "   [ -8  11 -14 ...  15  -6  12]]]\n",
      "\n",
      "\n",
      " ...\n",
      "\n",
      "\n",
      " [[[ -5   2  14 ...  -1  15   4]\n",
      "   [  2 -11  12 ...   4  11 -11]\n",
      "   [  3  -2  -9 ...  11  15 -12]\n",
      "   ...\n",
      "   [-15 -13  -4 ...  -2  -7 -13]\n",
      "   [  2 -15  10 ...  11  -2  -1]\n",
      "   [-12   8  -3 ...   0 -14  -3]]\n",
      "\n",
      "  [[  8  15  10 ...   8   9  -1]\n",
      "   [ -5   6  13 ...  13  13   5]\n",
      "   [-15  -7 -14 ...  -4  -7   5]\n",
      "   ...\n",
      "   [  5   0   4 ...   4  15  -3]\n",
      "   [ -2  16   1 ...  -2  -7   0]\n",
      "   [ -4   1  15 ...  -4  10 -16]]\n",
      "\n",
      "  [[-16   5   1 ...  -6   1   7]\n",
      "   [ 12  -1  -4 ...  10  16  -7]\n",
      "   [-14   9  16 ...  16  -4   1]\n",
      "   ...\n",
      "   [ 16   8  -6 ...  -5  -8  13]\n",
      "   [-14 -10   9 ... -10  12  -9]\n",
      "   [  9 -13   8 ...   3  15 -10]]]\n",
      "\n",
      "\n",
      " [[[ -3  14   2 ...  10  -3 -15]\n",
      "   [-11 -13  11 ... -16  10 -13]\n",
      "   [ -7 -14 -10 ...  -8 -15   9]\n",
      "   ...\n",
      "   [ -6   4  13 ...  -3   1  10]\n",
      "   [  5   6  -2 ...   2  15   8]\n",
      "   [  1  12  -7 ...   9  16  16]]\n",
      "\n",
      "  [[-13  16 -13 ... -15 -13  -2]\n",
      "   [ -2  -1  -6 ... -13 -11  -4]\n",
      "   [ -4  12  15 ...  11  -7  13]\n",
      "   ...\n",
      "   [ -7  13 -10 ...  10  -8 -10]\n",
      "   [ -7 -16  13 ...  -7  13 -12]\n",
      "   [ 15  11  -7 ...   0  16  -1]]\n",
      "\n",
      "  [[ -3   5  -2 ...  13 -13  16]\n",
      "   [ -3 -16  -3 ...  -6  -2  -9]\n",
      "   [-16  -6   9 ... -13  -6   7]\n",
      "   ...\n",
      "   [ 13   0   8 ...  15  -1  16]\n",
      "   [ -8   9  -3 ...  -1  12  -2]\n",
      "   [  2  10   6 ...   4   4 -12]]]\n",
      "\n",
      "\n",
      " [[[ -8 -15   1 ...   2  -2  11]\n",
      "   [ -3   6  15 ...   3  10 -11]\n",
      "   [  5  -5  -5 ...  -9  -5   4]\n",
      "   ...\n",
      "   [ -9  -8 -10 ... -12   8 -16]\n",
      "   [  3 -13  -5 ... -10  11   3]\n",
      "   [ -4  12  16 ...   9  13  -3]]\n",
      "\n",
      "  [[ -7  10  -3 ...  12  -8  -5]\n",
      "   [-14 -11   1 ...  -8  12  16]\n",
      "   [-12   4  -3 ... -15  15  10]\n",
      "   ...\n",
      "   [  9   9 -13 ... -15  -8 -10]\n",
      "   [ -3  14   6 ...  -1  -6  -9]\n",
      "   [-10   3  -9 ... -11   0  11]]\n",
      "\n",
      "  [[ -8   0 -11 ...  -9  12 -10]\n",
      "   [-13 -14 -13 ...  -3  16  -9]\n",
      "   [ -5  13 -11 ... -11  -5   0]\n",
      "   ...\n",
      "   [  2  -6 -10 ...  14  11   8]\n",
      "   [ 10 -16   2 ... -11  -9   2]\n",
      "   [  8  -4   4 ...  12  -7   3]]]]\n",
      "Layer: linear_pred_00\n",
      "- Weights:\n",
      "[[ 1  1  2 ...  0  2 -2]\n",
      " [-4 -4 -1 ... -4 -2 -4]\n",
      " [ 4 -4 -2 ...  2  0 -4]\n",
      " ...\n",
      " [ 1 -4 -1 ...  4 -4  3]\n",
      " [ 2  1 -4 ...  4  2 -3]\n",
      " [-5  5  3 ...  1  0  4]]\n",
      "Layer: conv_encoder_02\n",
      "- Weights:\n",
      "[[[[-5  6  1  7  1]\n",
      "   [ 3 -3  5  5 -3]\n",
      "   [-1  3  1  4  4]\n",
      "   [ 0 -1 -1  2  5]\n",
      "   [ 0  7 -7 -2  5]]\n",
      "\n",
      "  [[ 0  6 -3  5  0]\n",
      "   [-5  4  1 -4 -1]\n",
      "   [ 1 -3  0  3 -4]\n",
      "   [ 5 -6  7  5 -6]\n",
      "   [ 6  4 -1  6  5]]\n",
      "\n",
      "  [[ 1  7  1 -3 -3]\n",
      "   [ 5 -4  0  2  4]\n",
      "   [ 1  1 -7  5 -7]\n",
      "   [-6 -6 -3 -5 -7]\n",
      "   [-2  1  7  7 -6]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 2 -5  1  1  5]\n",
      "   [ 0 -5  0  3 -7]\n",
      "   [-3 -2  1 -6 -5]\n",
      "   [-4 -5 -4  0 -4]\n",
      "   [-5 -3  6 -7 -1]]\n",
      "\n",
      "  [[ 7 -2 -2 -2 -1]\n",
      "   [ 6  6 -1 -4  6]\n",
      "   [ 5 -5  0 -1 -4]\n",
      "   [ 4  2  5  0 -1]\n",
      "   [ 2 -1 -2  2  4]]\n",
      "\n",
      "  [[ 4 -5  5 -6  3]\n",
      "   [-2  2 -4  6  0]\n",
      "   [ 6  4  0  0  6]\n",
      "   [ 5  5  1  2  1]\n",
      "   [ 5  2 -6 -2 -6]]]\n",
      "\n",
      "\n",
      " [[[-6  6 -6 -1  2]\n",
      "   [-1 -5  0  1 -6]\n",
      "   [ 7 -3  3  7  7]\n",
      "   [ 1 -6 -7  2  2]\n",
      "   [-7  6  4 -2  7]]\n",
      "\n",
      "  [[ 4  3  4  0  3]\n",
      "   [-2  3  4  0  7]\n",
      "   [-4  4 -5 -2  1]\n",
      "   [-5  0  2 -1  5]\n",
      "   [-2 -4  6  1 -5]]\n",
      "\n",
      "  [[-1 -3 -7 -6  6]\n",
      "   [ 7 -3  1 -6 -1]\n",
      "   [ 0 -4  0  4 -1]\n",
      "   [-2  2 -6 -5  0]\n",
      "   [ 5  6  4 -6  4]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 7  3 -3  5 -1]\n",
      "   [-2  0  6  4 -6]\n",
      "   [ 7  0  6 -6  5]\n",
      "   [-1  7 -4 -7 -1]\n",
      "   [-3  6  5  5 -1]]\n",
      "\n",
      "  [[-6  7 -1  7 -4]\n",
      "   [ 1  5 -5  0 -6]\n",
      "   [ 3 -3 -5 -5 -7]\n",
      "   [-1  5 -1  4 -6]\n",
      "   [ 6 -3  1 -4 -3]]\n",
      "\n",
      "  [[-1  2 -2 -2  4]\n",
      "   [ 7  0 -2  2  3]\n",
      "   [ 0 -5 -1  3 -1]\n",
      "   [-1  6 -1 -1  7]\n",
      "   [-4 -1  4 -5  0]]]\n",
      "\n",
      "\n",
      " [[[-4 -1  3 -5  7]\n",
      "   [ 7  2 -1 -6  2]\n",
      "   [-7 -3 -3  5  1]\n",
      "   [ 3  4  1 -1 -3]\n",
      "   [-2  1  7 -2 -6]]\n",
      "\n",
      "  [[-4  4  5  6 -6]\n",
      "   [ 2 -6  2 -1  4]\n",
      "   [ 3 -3  4  2  0]\n",
      "   [-6  1  5 -6 -7]\n",
      "   [-4 -6 -7 -6 -6]]\n",
      "\n",
      "  [[ 2  4 -7 -2  2]\n",
      "   [ 5  0 -5 -5  7]\n",
      "   [-1  7 -7 -1  2]\n",
      "   [-4  7 -2  1  2]\n",
      "   [-2  1 -3  6  3]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-7  7 -4 -6 -4]\n",
      "   [-6  2  6  1  4]\n",
      "   [ 0  5  1 -2  3]\n",
      "   [ 2 -5 -5 -1 -6]\n",
      "   [-4  3 -2 -1 -2]]\n",
      "\n",
      "  [[-4 -7  7  6 -7]\n",
      "   [-4 -4  1  2 -6]\n",
      "   [ 3 -4  2  6 -3]\n",
      "   [ 5  1 -6  0  0]\n",
      "   [-1 -7  1  6  1]]\n",
      "\n",
      "  [[ 4 -7  1  4  5]\n",
      "   [-2 -5 -2  6 -7]\n",
      "   [ 0  6  0 -4 -5]\n",
      "   [-3 -2 -3 -7 -4]\n",
      "   [ 1  3 -5 -3  2]]]\n",
      "\n",
      "\n",
      " ...\n",
      "\n",
      "\n",
      " [[[ 4  6 -2 -6  6]\n",
      "   [-6  7  3  1 -2]\n",
      "   [-7 -5 -3  4 -2]\n",
      "   [-6 -6  1 -3 -1]\n",
      "   [-6  4 -1 -6  3]]\n",
      "\n",
      "  [[-4  5  5  1 -6]\n",
      "   [ 0  2  2 -3  6]\n",
      "   [-4 -6  3 -4  1]\n",
      "   [-3 -3  3  0  5]\n",
      "   [ 0 -7 -7  2 -5]]\n",
      "\n",
      "  [[-5  4  7 -7 -1]\n",
      "   [-7  1 -3 -4  4]\n",
      "   [-3 -1  2  6  1]\n",
      "   [ 5  4  0  6 -3]\n",
      "   [-5  0  5  3 -6]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-6  3  7 -2 -7]\n",
      "   [-6  5  6  0  6]\n",
      "   [-3 -7  0 -7  6]\n",
      "   [ 6 -1  0 -7 -1]\n",
      "   [ 0 -1  5 -7  5]]\n",
      "\n",
      "  [[ 7 -7  7 -1 -6]\n",
      "   [ 0 -7  6 -2  1]\n",
      "   [ 2  1 -7 -1  7]\n",
      "   [ 5 -6  7 -7 -1]\n",
      "   [-4  5 -1 -7 -1]]\n",
      "\n",
      "  [[ 6 -1 -3 -6 -7]\n",
      "   [ 5 -3  2  7 -6]\n",
      "   [ 4  5 -1 -1  4]\n",
      "   [-5  4  4  4  3]\n",
      "   [ 0 -6 -2 -5 -2]]]\n",
      "\n",
      "\n",
      " [[[-2  4  5 -2  2]\n",
      "   [-5 -6 -5 -4  1]\n",
      "   [ 5 -1 -1 -4  7]\n",
      "   [-4  7  0 -5  6]\n",
      "   [-5  5  2  6 -2]]\n",
      "\n",
      "  [[ 0 -6  2  1 -7]\n",
      "   [-7  5  3  4  7]\n",
      "   [ 4  1  7 -7  3]\n",
      "   [-2  1 -1  4  3]\n",
      "   [-4  7 -4  6 -6]]\n",
      "\n",
      "  [[ 6 -2  1  6  5]\n",
      "   [-7  2 -5 -4  7]\n",
      "   [-7  5  5  0  5]\n",
      "   [ 5 -6  6  3 -2]\n",
      "   [-7  1  0 -2  2]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-7 -7 -7 -5 -4]\n",
      "   [-2 -4 -7  1  0]\n",
      "   [-2 -6  4 -3  5]\n",
      "   [ 2  6 -3 -5 -7]\n",
      "   [ 2 -1 -5  6 -4]]\n",
      "\n",
      "  [[ 1  0  7  0  6]\n",
      "   [ 3 -3  5  2  1]\n",
      "   [-2  1  4  0  6]\n",
      "   [ 1 -4  5  3 -5]\n",
      "   [ 7 -4  4 -6  0]]\n",
      "\n",
      "  [[-2  2  5  4  4]\n",
      "   [-6  4  3 -2 -4]\n",
      "   [ 5 -5 -1  5 -1]\n",
      "   [ 7 -7  5 -4 -1]\n",
      "   [ 2 -1  5 -5  1]]]\n",
      "\n",
      "\n",
      " [[[ 1  1 -7  0 -2]\n",
      "   [ 3  6 -2  1 -4]\n",
      "   [ 6  1  7  5  3]\n",
      "   [-1 -1 -3 -6  7]\n",
      "   [-4  3  2  5  3]]\n",
      "\n",
      "  [[ 0 -4  1  0  6]\n",
      "   [-4 -1  3  6 -6]\n",
      "   [ 3  1  7 -5  6]\n",
      "   [ 0  5 -7  6 -5]\n",
      "   [-2  3  3 -7 -2]]\n",
      "\n",
      "  [[ 5  6  2 -2 -7]\n",
      "   [ 4  1  7  7 -3]\n",
      "   [-5  1 -5 -5  5]\n",
      "   [ 3 -4  7 -5 -3]\n",
      "   [-7 -3  5  7  5]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-6 -7  1 -3 -3]\n",
      "   [ 5  1  2  0 -5]\n",
      "   [ 0 -6 -7 -3 -6]\n",
      "   [ 3 -4  6  1 -4]\n",
      "   [ 1  6 -5  0 -7]]\n",
      "\n",
      "  [[-5 -7 -4  2  1]\n",
      "   [ 6  1  3 -6  4]\n",
      "   [ 7  0 -2  4  2]\n",
      "   [ 0 -7 -4  3  2]\n",
      "   [ 5 -7  4 -2 -6]]\n",
      "\n",
      "  [[-3  7 -6 -5 -4]\n",
      "   [-5 -6  6  1  7]\n",
      "   [-6  6 -7 -4  6]\n",
      "   [ 5  0  7  6  5]\n",
      "   [-2 -5  1  3 -7]]]]\n",
      "Layer: linear_pred_02\n",
      "- Weights:\n",
      "[[ 0  0 -4 ...  3 -2 -2]\n",
      " [ 0 -3  1 ... -2  5 -4]\n",
      " [-1  0 -4 ...  2 -5 -5]\n",
      " ...\n",
      " [ 4  1 -5 ... -1  0 -1]\n",
      " [ 2  4 -2 ...  5  3  1]\n",
      " [-4  0  3 ... -1  4  0]]\n",
      "Layer: conv_encoder_04\n",
      "- Weights:\n",
      "[[[[-4 -4  5]\n",
      "   [-3  6 -7]\n",
      "   [-7 -7  5]]\n",
      "\n",
      "  [[ 0 -1 -4]\n",
      "   [-6  1 -7]\n",
      "   [ 5  1 -3]]\n",
      "\n",
      "  [[-1  4 -5]\n",
      "   [-6 -4 -7]\n",
      "   [ 3  7  0]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 0 -1  6]\n",
      "   [ 2 -2  4]\n",
      "   [ 1 -4 -7]]\n",
      "\n",
      "  [[-6  1  0]\n",
      "   [-6 -5  1]\n",
      "   [ 4  2 -4]]\n",
      "\n",
      "  [[-3 -1  7]\n",
      "   [-3  2 -6]\n",
      "   [ 4  2 -1]]]\n",
      "\n",
      "\n",
      " [[[ 1 -2  7]\n",
      "   [-4 -3  6]\n",
      "   [-4  1 -2]]\n",
      "\n",
      "  [[ 2  5  6]\n",
      "   [ 5 -2 -5]\n",
      "   [-3  0 -3]]\n",
      "\n",
      "  [[ 3 -7  2]\n",
      "   [ 6 -1  3]\n",
      "   [-3  4 -3]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 6  4 -4]\n",
      "   [ 4  3 -1]\n",
      "   [ 0  5  0]]\n",
      "\n",
      "  [[ 5  0  1]\n",
      "   [-3 -1 -7]\n",
      "   [ 1 -2 -4]]\n",
      "\n",
      "  [[-6 -4 -4]\n",
      "   [-1 -6 -7]\n",
      "   [-2  0 -5]]]\n",
      "\n",
      "\n",
      " [[[ 0  3 -5]\n",
      "   [-6 -2  3]\n",
      "   [-1 -7 -7]]\n",
      "\n",
      "  [[ 5 -4  3]\n",
      "   [ 5  2  2]\n",
      "   [-1  7  3]]\n",
      "\n",
      "  [[ 1  7  5]\n",
      "   [ 0  5  2]\n",
      "   [-4 -4 -6]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 1 -3  1]\n",
      "   [ 1  4 -6]\n",
      "   [-5  1  6]]\n",
      "\n",
      "  [[-4  5 -4]\n",
      "   [ 0 -4 -4]\n",
      "   [-7 -1 -5]]\n",
      "\n",
      "  [[ 3  2  4]\n",
      "   [-5  2  3]\n",
      "   [ 3  0  4]]]\n",
      "\n",
      "\n",
      " ...\n",
      "\n",
      "\n",
      " [[[-1  1  0]\n",
      "   [-4  3 -4]\n",
      "   [ 6 -1  0]]\n",
      "\n",
      "  [[ 7 -4 -6]\n",
      "   [ 1  5  5]\n",
      "   [ 1 -2  0]]\n",
      "\n",
      "  [[-2  3  2]\n",
      "   [-2  5 -5]\n",
      "   [-1  7  6]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 7 -3  6]\n",
      "   [-3 -5  7]\n",
      "   [ 0 -4  1]]\n",
      "\n",
      "  [[-6 -5 -5]\n",
      "   [-5  7 -3]\n",
      "   [-7  0  2]]\n",
      "\n",
      "  [[ 4 -4 -6]\n",
      "   [ 6  5  4]\n",
      "   [-6 -3 -5]]]\n",
      "\n",
      "\n",
      " [[[-5  1 -1]\n",
      "   [ 6 -1  0]\n",
      "   [ 2 -5  6]]\n",
      "\n",
      "  [[ 5 -3  3]\n",
      "   [ 2  0  6]\n",
      "   [ 5 -5 -1]]\n",
      "\n",
      "  [[-1 -6  3]\n",
      "   [-4 -5  7]\n",
      "   [-5  4 -6]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-3  3  5]\n",
      "   [-3  2 -5]\n",
      "   [-7  3  5]]\n",
      "\n",
      "  [[-1 -2 -2]\n",
      "   [ 5  3 -6]\n",
      "   [ 5 -7  6]]\n",
      "\n",
      "  [[ 7 -7  6]\n",
      "   [ 1 -4 -6]\n",
      "   [ 0  3 -7]]]\n",
      "\n",
      "\n",
      " [[[-6  7 -3]\n",
      "   [-6  1  6]\n",
      "   [-1 -5 -1]]\n",
      "\n",
      "  [[ 2  1  5]\n",
      "   [ 2 -4  0]\n",
      "   [ 5 -4  2]]\n",
      "\n",
      "  [[-3  3  7]\n",
      "   [ 0  5  0]\n",
      "   [-7 -6  2]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 4  2  3]\n",
      "   [ 2  6 -2]\n",
      "   [-2 -2 -6]]\n",
      "\n",
      "  [[-6  0  1]\n",
      "   [ 0  7 -1]\n",
      "   [-5 -4  4]]\n",
      "\n",
      "  [[ 5 -6 -3]\n",
      "   [-7  1  7]\n",
      "   [ 6 -3  7]]]]\n",
      "Layer: linear_pred_04\n",
      "- Weights:\n",
      "[[ 5  4  0 ...  1 -5  5]\n",
      " [ 0 -3  1 ... -2 -2  4]\n",
      " [-1  1  0 ... -2 -1 -4]\n",
      " ...\n",
      " [ 2 -5 -1 ...  4  4 -1]\n",
      " [-3  0 -5 ... -5  5 -2]\n",
      " [-3  3 -2 ... -1  1  0]]\n",
      "Layer: conv_encoder_05\n",
      "- Weights:\n",
      "[[[[ 4  2  0]\n",
      "   [-4 -2  3]\n",
      "   [ 5 -5  4]]\n",
      "\n",
      "  [[-2 -3 -4]\n",
      "   [ 4 -2 -4]\n",
      "   [ 1  0 -5]]\n",
      "\n",
      "  [[-2 -5  1]\n",
      "   [-2 -2  2]\n",
      "   [ 5 -5  2]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 4  4 -1]\n",
      "   [-3 -4  2]\n",
      "   [ 0  1  1]]\n",
      "\n",
      "  [[ 1  4  0]\n",
      "   [ 2  5  1]\n",
      "   [ 5 -5 -4]]\n",
      "\n",
      "  [[ 4 -4  5]\n",
      "   [-3 -1 -2]\n",
      "   [-1  0 -2]]]\n",
      "\n",
      "\n",
      " [[[-2  3 -3]\n",
      "   [-2 -5 -2]\n",
      "   [ 5 -4  1]]\n",
      "\n",
      "  [[-3  4 -2]\n",
      "   [-3 -1  5]\n",
      "   [ 2 -2  4]]\n",
      "\n",
      "  [[-3 -1 -1]\n",
      "   [ 2  2 -2]\n",
      "   [-3  3  1]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 1 -5 -2]\n",
      "   [-4  2 -3]\n",
      "   [-2 -1 -4]]\n",
      "\n",
      "  [[ 3 -3  5]\n",
      "   [-3 -1 -5]\n",
      "   [-4 -3 -1]]\n",
      "\n",
      "  [[-4 -1 -3]\n",
      "   [-2  5  3]\n",
      "   [ 2 -4  1]]]\n",
      "\n",
      "\n",
      " [[[ 5 -3  3]\n",
      "   [-5 -4  3]\n",
      "   [ 3  2  3]]\n",
      "\n",
      "  [[-4  4  2]\n",
      "   [ 4  1  4]\n",
      "   [ 2 -4 -1]]\n",
      "\n",
      "  [[ 2  1 -2]\n",
      "   [-1  3  2]\n",
      "   [ 2 -5  2]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 3 -2 -2]\n",
      "   [-5  1 -5]\n",
      "   [ 0  1 -1]]\n",
      "\n",
      "  [[ 3 -4 -3]\n",
      "   [ 4  4  1]\n",
      "   [ 2  5  1]]\n",
      "\n",
      "  [[-1  0 -5]\n",
      "   [-4 -2 -1]\n",
      "   [ 2 -1  0]]]\n",
      "\n",
      "\n",
      " ...\n",
      "\n",
      "\n",
      " [[[-4 -2 -2]\n",
      "   [ 4 -3 -3]\n",
      "   [ 5 -4 -4]]\n",
      "\n",
      "  [[-5 -3  2]\n",
      "   [-5  1  0]\n",
      "   [-1 -3  3]]\n",
      "\n",
      "  [[ 0 -4 -2]\n",
      "   [-3 -4  3]\n",
      "   [-2  1 -5]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 5  1 -4]\n",
      "   [-1 -2  2]\n",
      "   [ 3  3 -2]]\n",
      "\n",
      "  [[-4  3 -4]\n",
      "   [-4 -3 -2]\n",
      "   [ 1  1 -3]]\n",
      "\n",
      "  [[ 3  0  3]\n",
      "   [-5  2  1]\n",
      "   [ 0  2  1]]]\n",
      "\n",
      "\n",
      " [[[ 0  5 -1]\n",
      "   [ 2  2 -3]\n",
      "   [-2  3 -1]]\n",
      "\n",
      "  [[ 0  2  3]\n",
      "   [ 0 -2 -2]\n",
      "   [ 1 -4 -4]]\n",
      "\n",
      "  [[-5  3  5]\n",
      "   [-2 -3  3]\n",
      "   [ 2  2  0]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-3 -4  1]\n",
      "   [ 2 -2 -3]\n",
      "   [ 3 -3 -2]]\n",
      "\n",
      "  [[ 2  5 -5]\n",
      "   [ 2  4  5]\n",
      "   [-4  1 -3]]\n",
      "\n",
      "  [[-1 -5 -1]\n",
      "   [ 3 -3  5]\n",
      "   [-1  3  1]]]\n",
      "\n",
      "\n",
      " [[[-2  0  3]\n",
      "   [ 0 -4  4]\n",
      "   [-5  1 -3]]\n",
      "\n",
      "  [[-3  1 -5]\n",
      "   [-5  5 -1]\n",
      "   [-4 -5 -3]]\n",
      "\n",
      "  [[-2 -4  0]\n",
      "   [ 4 -5  2]\n",
      "   [ 5  5  1]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 5  2 -3]\n",
      "   [-5 -1 -2]\n",
      "   [ 4 -3 -4]]\n",
      "\n",
      "  [[ 4 -4  1]\n",
      "   [-2  3  1]\n",
      "   [-5  0 -3]]\n",
      "\n",
      "  [[ 4 -2 -5]\n",
      "   [ 4  1 -5]\n",
      "   [-3 -2  2]]]]\n",
      "Layer: linear_pred_05\n",
      "- Weights:\n",
      "[[ 0 -6 -1 ...  4  5  3]\n",
      " [ 4 -6 -1 ... -6 -2  4]\n",
      " [ 0  3  5 ... -4  2 -1]\n",
      " ...\n",
      " [ 6 -2  4 ...  6 -6  5]\n",
      " [ 6 -5 -1 ...  1 -2 -4]\n",
      " [-3  6 -5 ...  5 -5 -5]]\n",
      "Layer: conv_encoder_06\n",
      "- Weights:\n",
      "[[[[-3  2  4]\n",
      "   [-3 -2  4]\n",
      "   [ 3 -5  2]]\n",
      "\n",
      "  [[-4 -2  0]\n",
      "   [ 4  3 -3]\n",
      "   [-5  0  1]]\n",
      "\n",
      "  [[-3 -2  4]\n",
      "   [ 1  3 -4]\n",
      "   [ 0  5  6]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-1  5 -1]\n",
      "   [-3 -1 -2]\n",
      "   [-5  6 -2]]\n",
      "\n",
      "  [[-1 -5 -4]\n",
      "   [ 2 -1  4]\n",
      "   [-4  6 -6]]\n",
      "\n",
      "  [[ 5 -2 -1]\n",
      "   [ 6 -5 -3]\n",
      "   [ 2 -2  6]]]\n",
      "\n",
      "\n",
      " [[[ 0  4 -6]\n",
      "   [-6  5 -5]\n",
      "   [-3  0 -3]]\n",
      "\n",
      "  [[ 3  2 -4]\n",
      "   [ 1 -1  1]\n",
      "   [ 2  3 -2]]\n",
      "\n",
      "  [[-6 -2  0]\n",
      "   [-3  3  3]\n",
      "   [-2 -6 -1]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 1  0  1]\n",
      "   [ 0  6  3]\n",
      "   [ 5 -2 -2]]\n",
      "\n",
      "  [[ 5  3  1]\n",
      "   [ 1 -6 -4]\n",
      "   [-1 -1  2]]\n",
      "\n",
      "  [[-6  4  1]\n",
      "   [ 6 -4 -6]\n",
      "   [ 4 -1 -3]]]\n",
      "\n",
      "\n",
      " [[[ 3 -4  6]\n",
      "   [-4 -4  0]\n",
      "   [-5 -6 -4]]\n",
      "\n",
      "  [[-1 -2  5]\n",
      "   [ 0  0  6]\n",
      "   [ 4  4  6]]\n",
      "\n",
      "  [[ 6  4 -6]\n",
      "   [-4 -3 -2]\n",
      "   [-6 -5  1]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 0  1  1]\n",
      "   [-1  5 -2]\n",
      "   [-6  0  5]]\n",
      "\n",
      "  [[ 2  3 -1]\n",
      "   [-1 -5  0]\n",
      "   [-5  2 -3]]\n",
      "\n",
      "  [[ 1  3  3]\n",
      "   [-2 -5 -1]\n",
      "   [-4  6 -3]]]\n",
      "\n",
      "\n",
      " ...\n",
      "\n",
      "\n",
      " [[[-2  6  0]\n",
      "   [-4  3  1]\n",
      "   [ 4  2 -5]]\n",
      "\n",
      "  [[ 5  5 -3]\n",
      "   [ 1  2 -4]\n",
      "   [ 2  1 -5]]\n",
      "\n",
      "  [[-5 -4 -4]\n",
      "   [-6 -5  4]\n",
      "   [ 0 -2 -3]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-3 -4 -1]\n",
      "   [ 0 -2  6]\n",
      "   [ 1 -4 -6]]\n",
      "\n",
      "  [[-3  4 -3]\n",
      "   [ 0  1  2]\n",
      "   [-3  4  1]]\n",
      "\n",
      "  [[ 1 -6  3]\n",
      "   [-1  2 -2]\n",
      "   [-4 -3  0]]]\n",
      "\n",
      "\n",
      " [[[ 4  3 -1]\n",
      "   [-5  6  2]\n",
      "   [-2 -5 -1]]\n",
      "\n",
      "  [[ 1  2 -3]\n",
      "   [-5 -4 -1]\n",
      "   [ 2 -2  5]]\n",
      "\n",
      "  [[-5 -1 -3]\n",
      "   [-4  4 -2]\n",
      "   [-3 -1  5]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[ 2 -2  0]\n",
      "   [ 5  1  5]\n",
      "   [-3  4  6]]\n",
      "\n",
      "  [[ 6 -5  5]\n",
      "   [ 6 -6  3]\n",
      "   [ 5 -5  2]]\n",
      "\n",
      "  [[-4 -4  0]\n",
      "   [-2  6 -2]\n",
      "   [-2  3 -5]]]\n",
      "\n",
      "\n",
      " [[[-1  3  6]\n",
      "   [-5  5 -3]\n",
      "   [-6  1  1]]\n",
      "\n",
      "  [[ 5 -6 -2]\n",
      "   [ 6 -3 -5]\n",
      "   [ 4  6  2]]\n",
      "\n",
      "  [[ 2  3  0]\n",
      "   [ 5  6 -2]\n",
      "   [ 5  5 -1]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[-3  2  5]\n",
      "   [-6  5  6]\n",
      "   [ 4  2 -2]]\n",
      "\n",
      "  [[-2 -1  0]\n",
      "   [-5  6 -5]\n",
      "   [ 2 -1  0]]\n",
      "\n",
      "  [[-5  1 -2]\n",
      "   [ 4 -3 -2]\n",
      "   [-4  3  2]]]]\n",
      "Layer: linear_pred_06\n",
      "- Weights:\n",
      "[[-5 -1  6 ...  1  0 -6]\n",
      " [ 6  0 -5 ... -4  3  6]\n",
      " [-3 -3  3 ...  0  1  3]\n",
      " ...\n",
      " [-2  3  0 ...  2 -3  0]\n",
      " [ 5  5  2 ... -6  2 -6]\n",
      " [ 6  5 -1 ... -2 -6  1]]\n",
      "Layer: linear_encoder_08\n",
      "- Weights:\n",
      "[[ 1 -6  4 ...  3  4 -5]\n",
      " [-1 -4 -2 ...  2  6 -6]\n",
      " [ 4  6 -4 ...  0  1 -6]\n",
      " ...\n",
      " [-5  0  0 ... -6  0 -3]\n",
      " [-2 -6 -5 ... -2  4  3]\n",
      " [-4  1 -6 ... -1  5  2]]\n",
      "Layer: linear_pred_08\n",
      "- Weights:\n",
      "[[-7 -9 -7 ...  9  7 -5]\n",
      " [ 6 -7  4 ... -5  5 -6]\n",
      " [ 8  4  3 ...  9 -5  7]\n",
      " ...\n",
      " [ 2 -3  9 ... -5  1 -3]\n",
      " [ 2  3 -7 ...  5  8 -9]\n",
      " [-9  8  9 ...  0  3 -8]]\n",
      "Layer: linear_10\n",
      "- Weights:\n",
      "[[ 7 -3  4 ...  9  5 -8]\n",
      " [ 5 -5 -7 ... -2  1 -1]\n",
      " [-8  0  5 ...  9  7 -3]\n",
      " ...\n",
      " [ 9  9 -9 ... -8 -1  0]\n",
      " [-4  6  1 ... -9  1  4]\n",
      " [-1 -8 -5 ... -5  5  7]]\n"
     ]
    }
   ],
   "source": [
    "model.print_layers_parameters()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f85ea3d88bd78a9a",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "# Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "846c609e4fcdbd92",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:06.378159Z",
     "start_time": "2024-05-14T16:09:06.375462Z"
    }
   },
   "outputs": [],
   "source": [
    "from lib.optim.lr_schedulers import ConstantLR\n",
    "from lib.utils.nn import EarlyStopping, ReduceLROnPlateau"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "36729474350b812d",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-05-14T16:09:06.382558Z",
     "start_time": "2024-05-14T16:09:06.379168Z"
    },
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "lr_scheduler = ConstantLR(config['lr_inv'])\n",
    "early_stopping = EarlyStopping(min_delta=0.01, patience=25, from_epoch=10)\n",
    "reduce_lr_plateau = ReduceLROnPlateau(factor=2, min_delta=0.01, patience=15, from_epoch=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "17c0361566342239",
   "metadata": {
    "ExecuteTime": {
     "start_time": "2024-05-14T16:09:06.383568Z"
    },
    "collapsed": false,
    "jupyter": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "afaadf61c60547a8ad0e190daba46d0b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/35 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 0:\n",
      "- Local accuracy 0: 27.88%\n",
      "- Local accuracy 1: 9.51%\n",
      "- Local accuracy 2: 9.70%\n",
      "- Local accuracy 3: 9.86%\n",
      "- Local accuracy 4: 9.52%\n",
      "- Local accuracy 5: 14.98%\n",
      "- Output accuracy: 9.92%\n",
      "- Output val accuracy: 9.66%\n",
      "\n",
      "\n",
      "Epoch 3:\n",
      "- Local accuracy 0: 47.54%\n",
      "- Local accuracy 1: 39.50%\n",
      "- Local accuracy 2: 36.84%\n",
      "- Local accuracy 3: 34.44%\n",
      "- Local accuracy 4: 32.64%\n",
      "- Local accuracy 5: 20.37%\n",
      "- Output accuracy: 13.96%\n",
      "- Output val accuracy: 22.12%\n",
      "\n",
      "\n",
      "Epoch 6:\n",
      "- Local accuracy 0: 51.64%\n",
      "- Local accuracy 1: 47.69%\n",
      "- Local accuracy 2: 45.07%\n",
      "- Local accuracy 3: 42.55%\n",
      "- Local accuracy 4: 41.14%\n",
      "- Local accuracy 5: 39.06%\n",
      "- Output accuracy: 34.01%\n",
      "- Output val accuracy: 39.71%\n",
      "\n",
      "\n",
      "Epoch 9:\n",
      "- Local accuracy 0: 53.55%\n",
      "- Local accuracy 1: 52.32%\n",
      "- Local accuracy 2: 50.67%\n",
      "- Local accuracy 3: 48.79%\n",
      "- Local accuracy 4: 47.33%\n",
      "- Local accuracy 5: 47.35%\n",
      "- Output accuracy: 44.62%\n",
      "- Output val accuracy: 44.79%\n",
      "\n",
      "\n",
      "Epoch 12:\n",
      "- Local accuracy 0: 55.15%\n",
      "- Local accuracy 1: 55.20%\n",
      "- Local accuracy 2: 54.27%\n",
      "- Local accuracy 3: 52.51%\n",
      "- Local accuracy 4: 51.39%\n",
      "- Local accuracy 5: 51.76%\n",
      "- Output accuracy: 49.97%\n",
      "- Output val accuracy: 45.12%\n",
      "\n",
      "\n",
      "Epoch 15:\n",
      "- Local accuracy 0: 55.30%\n",
      "- Local accuracy 1: 57.21%\n",
      "- Local accuracy 2: 56.86%\n",
      "- Local accuracy 3: 55.03%\n",
      "- Local accuracy 4: 53.88%\n",
      "- Local accuracy 5: 53.92%\n",
      "- Output accuracy: 52.74%\n",
      "- Output val accuracy: 56.81%\n",
      "\n",
      "\n",
      "Epoch 18:\n",
      "- Local accuracy 0: 56.38%\n",
      "- Local accuracy 1: 59.29%\n",
      "- Local accuracy 2: 59.35%\n",
      "- Local accuracy 3: 57.32%\n",
      "- Local accuracy 4: 56.46%\n",
      "- Local accuracy 5: 56.58%\n",
      "- Output accuracy: 55.68%\n",
      "- Output val accuracy: 58.78%\n",
      "\n",
      "\n",
      "Epoch 21:\n",
      "- Local accuracy 0: 56.80%\n",
      "- Local accuracy 1: 60.72%\n",
      "- Local accuracy 2: 61.05%\n",
      "- Local accuracy 3: 59.18%\n",
      "- Local accuracy 4: 58.35%\n",
      "- Local accuracy 5: 58.18%\n",
      "- Output accuracy: 57.62%\n",
      "- Output val accuracy: 59.41%\n",
      "\n",
      "\n",
      "Epoch 24:\n",
      "- Local accuracy 0: 57.63%\n",
      "- Local accuracy 1: 62.66%\n",
      "- Local accuracy 2: 63.34%\n",
      "- Local accuracy 3: 61.71%\n",
      "- Local accuracy 4: 60.86%\n",
      "- Local accuracy 5: 60.68%\n",
      "- Output accuracy: 60.43%\n",
      "- Output val accuracy: 61.83%\n",
      "\n",
      "\n",
      "Epoch 27:\n",
      "- Local accuracy 0: 57.98%\n",
      "- Local accuracy 1: 63.36%\n",
      "- Local accuracy 2: 64.45%\n",
      "- Local accuracy 3: 62.90%\n",
      "- Local accuracy 4: 61.81%\n",
      "- Local accuracy 5: 61.79%\n",
      "- Output accuracy: 61.35%\n",
      "- Output val accuracy: 59.89%\n",
      "\n",
      "\n",
      "Epoch 30:\n",
      "- Local accuracy 0: 58.90%\n",
      "- Local accuracy 1: 64.93%\n",
      "- Local accuracy 2: 66.07%\n",
      "- Local accuracy 3: 64.72%\n",
      "- Local accuracy 4: 63.83%\n",
      "- Local accuracy 5: 63.63%\n",
      "- Output accuracy: 63.07%\n",
      "- Output val accuracy: 64.23%\n",
      "\n",
      "\n",
      "Epoch 33:\n",
      "- Local accuracy 0: 58.47%\n",
      "- Local accuracy 1: 65.43%\n",
      "- Local accuracy 2: 66.63%\n",
      "- Local accuracy 3: 65.31%\n",
      "- Local accuracy 4: 64.56%\n",
      "- Local accuracy 5: 64.42%\n",
      "- Output accuracy: 64.02%\n",
      "- Output val accuracy: 63.93%\n",
      "\n",
      "\n",
      "Final train loss: 3928.67949\n",
      "Final train accuracy: 64.43%\n",
      "Final validation loss: 3911.39103\n",
      "Final validation accuracy: 64.95%\n"
     ]
    }
   ],
   "source": [
    "n_train_batches = len(X_train) // config['batch_size']\n",
    "n_val_batches = len(X_val) // config['batch_size']\n",
    "last_epoch = 0\n",
    "\n",
    "xp = cp.get_array_module(X_train)\n",
    "loss_history = xp.empty(config['num_epochs'])\n",
    "acc_history = xp.empty(config['num_epochs'])\n",
    "val_loss_history = xp.empty(config['num_epochs'])\n",
    "val_acc_history = xp.empty(config['num_epochs'])\n",
    "\n",
    "for epoch in trange(config['num_epochs']):\n",
    "    # Train and validation loops\n",
    "    lr_scheduler.step()\n",
    "    X_train, y_train = shuffle_dataset(X_train, y_train)\n",
    "    \n",
    "    loss_history[epoch], acc_history[epoch], local_losses, local_accuracies = train_step_local(\n",
    "        model=model,\n",
    "        X_train=X_train,\n",
    "        y_train=y_train,\n",
    "        batch_size=config['batch_size'],\n",
    "        n_train_batches=n_train_batches,\n",
    "        augmentation_pipeline=augmentation_pipeline,\n",
    "        lr_inv=lr_scheduler.get_lr()\n",
    "    )\n",
    "    \n",
    "    val_loss_history[epoch], val_acc_history[epoch] = val_step(\n",
    "        model=model, \n",
    "        X_val=X_val, \n",
    "        y_val=y_val, \n",
    "        batch_size=config['batch_size'], \n",
    "        n_val_batches=n_val_batches\n",
    "    )\n",
    "    \n",
    "    last_epoch = epoch\n",
    "\n",
    "    if epoch % 3 == 0:\n",
    "        print(f\"Epoch {epoch}:\")\n",
    "        for i, (local_loss, local_acc) in enumerate(zip(local_losses, local_accuracies)):\n",
    "            print(f\"- Local accuracy {i}: {local_acc:.2%}\")\n",
    "        print(f\"- Output accuracy: {acc_history[epoch]:.2%}\")\n",
    "        print(f\"- Output val accuracy: {val_acc_history[epoch]:.2%}\")\n",
    "        print(\"\\n\")\n",
    "\n",
    "    # Early stopping\n",
    "        if early_stopping.early_stop(float(val_acc_history[epoch])):\n",
    "            print(f\"{'-' * 64}\\nEarly stopping at epoch {epoch + 1}\\n{'-' * 64}\")\n",
    "            break\n",
    "\n",
    "    # ReduceLROnPlateau\n",
    "    if reduce_lr_plateau.reduce_lr(float(val_acc_history[epoch])):\n",
    "        lr_scheduler.set_lr(lr_scheduler.get_lr() * reduce_lr_plateau.factor)\n",
    "        print(f\"{'-' * 64}\\nReducing learning rate on plateau at epoch {epoch + 1}\\n{'-' * 64}\")\n",
    "    \n",
    "print(f'Final train loss: {loss_history[last_epoch]:.5f}')\n",
    "print(f'Final train accuracy: {acc_history[last_epoch]:.2%}')\n",
    "print(f'Final validation loss: {val_loss_history[last_epoch]:.5f}')\n",
    "print(f'Final validation accuracy: {val_acc_history[last_epoch]:.2%}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "58908f88838a9604",
   "metadata": {
    "jupyter": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Local Val Accuracy 0: 61.20%\n",
      "Local Val Accuracy 1: 65.54%\n",
      "Local Val Accuracy 2: 66.69%\n",
      "Local Val Accuracy 3: 66.09%\n",
      "Local Val Accuracy 4: 64.94%\n",
      "Local Val Accuracy 5: 65.23%\n"
     ]
    }
   ],
   "source": [
    "y_subnets = model.subnetworks_predict(X_test)\n",
    "for i, y_subnet in enumerate(y_subnets):\n",
    "    subnet_acc = accuracy(y_test, y_subnet)\n",
    "    print(f'Local Val Accuracy {i}: {subnet_acc:.2%}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "fe3a69e1b0645ef3",
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "loss_history, val_loss_history  = loss_history[:last_epoch+1], val_loss_history[:last_epoch+1]\n",
    "acc_history, val_acc_history = acc_history[:last_epoch+1], val_acc_history[:last_epoch+1]\n",
    "\n",
    "if config['device'] == 'cuda':\n",
    "    # Move training statistics to the CPU\n",
    "    loss_history, val_loss_history = to_cpu(loss_history), to_cpu(val_loss_history)\n",
    "    acc_history, val_acc_history = to_cpu(acc_history), to_cpu(val_acc_history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "8b3c8ec61925c880",
   "metadata": {
    "collapsed": false,
    "jupyter": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1600x500 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_history(loss_history, val_loss_history, acc_history, val_acc_history, log_scale=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "8662b1f12e958d1c",
   "metadata": {
    "jupyter": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test accuracy: 65.19%\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict(X_test, progress_bar=False)\n",
    "test_loss = l2_loss(y_test, y_pred)\n",
    "test_acc = accuracy(y_test, y_pred)\n",
    "print(f'Test accuracy: {test_acc:.2%}')"
   ]
  }
 ],
 "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": "3.11.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}