getGradients.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import argparse\n",
    "import os\n",
    "import random\n",
    "import shutil\n",
    "import time\n",
    "import warnings\n",
    "import json\n",
    "import re\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.parallel\n",
    "import torch.backends.cudnn as cudnn\n",
    "import torch.distributed as dist\n",
    "import torch.optim\n",
    "import torch.utils.data\n",
    "import torch.utils.data.distributed\n",
    "from torch.utils.data.sampler import SubsetRandomSampler\n",
    "from torchvision import datasets, transforms\n",
    "\n",
    "from model_analysis_nets import *\n",
    "from CKA import linear_CKA, kernel_CKA\n",
    "\n",
    "import numpy as np\n",
    "import sympy\n",
    "import pandas\n",
    "import seaborn\n",
    "\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def getGradients(modelpth : str = \"\"):\n",
    "    if not modelpth:\n",
    "        return {}\n",
    "    model1 = LeNet().to(device)\n",
    "    model1.load_state_dict(torch.load(modelpth))\n",
    "    \n",
    "    criterion = nn.CrossEntropyLoss()\n",
    "    optimizer = torch.optim.SGD(model1.parameters(), lr=0.01, momentum=0.5)\n",
    "\n",
    "    for X, Y in test_loader:\n",
    "        X_test = X.to(device)\n",
    "        Y_test = Y.to(device)\n",
    "        # print(Y)\n",
    "        break\n",
    "\n",
    "    # Forward pass\n",
    "    outputs = model1(X_test)\n",
    "\n",
    "    # Compute the loss\n",
    "    loss = criterion(outputs, Y_test)\n",
    "\n",
    "    # Zero the gradients\n",
    "    optimizer.zero_grad()\n",
    "\n",
    "    # Backward pass\n",
    "    loss.backward()\n",
    "\n",
    "    # Access the gradients\n",
    "    gradients1 = {}\n",
    "    for name, param in model1.named_parameters():\n",
    "        gradients1[name] = param.grad\n",
    "    return gradients1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "trans_mnist = transforms.Compose([transforms.ToTensor(),\n",
    "                                  transforms.Normalize((0.1307,), (0.3081,))])\n",
    "dataset_test = datasets.MNIST('data/mnist/', train=False, download=True, transform=trans_mnist)\n",
    "test_loader = torch.utils.data.DataLoader(\n",
    "                dataset_test, batch_size=32,\n",
    "                num_workers=2, pin_memory=True, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "device = 'cuda' if torch.cuda.is_available() else 'cpu'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# attack_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/cifar10/vgg_iidTrue_num100_C0.8_le2/shard2/pattern11-30--17-48-57/local_attack_save\"\n",
    "# normal_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/cifar10/vgg_iidTrue_num100_C0.8_le2/shard2/pattern11-30--17-48-57/local_normal_save\"\n",
    "attack_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/mnist/lenet_iidTrue_num100_C0.8_le2/shard2/pattern11-30--17-49-20/local_attack_save/\"\n",
    "normal_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/mnist/lenet_iidTrue_num100_C0.8_le2/shard2/pattern11-30--17-49-20/local_normal_save/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# md1_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/cifar10/resnet20_iidTrue_num100_C0.8_le2/shard2/pattern12-01--00-32-01/local_attack_save/iter_18_attack_0.pt\"\n",
    "# md2_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/cifar10/resnet20_iidTrue_num100_C0.8_le2/shard2/pattern12-01--00-32-01/local_normal_save/iter_18_normal_21.pt\"\n",
    "# md1_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/mnist/lenet_iidTrue_num100_C0.8_le2/shard2/pattern11-30--17-49-20/local_attack_save/iter_8_attack_5.pt\"\n",
    "# md2_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/mnist/lenet_iidTrue_num100_C0.8_le2/shard2/pattern11-30--17-49-20/local_normal_save/iter_8_normal_24.pt\"\n",
    "# md3_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/mnist/lenet_iidTrue_num100_C0.8_le2/shard2/pattern11-30--17-49-20/local_attack_save/iter_8_attack_2.pt\"\n",
    "# md4_pth = \"../TDA-NN/3090/LG-FedAvg/save_attack_ub/mnist/lenet_iidTrue_num100_C0.8_le2/shard2/pattern11-30--17-49-20/local_normal_save/iter_8_normal_27.pt\"\n",
    "\n",
    "# md1 = torch.load(md1_pth)\n",
    "# md2 = torch.load(md2_pth)\n",
    "# md3 = torch.load(md3_pth)\n",
    "# md4 = torch.load(md4_pth)\n",
    "# model1 = LeNet().to(device)\n",
    "# model1.load_state_dict(torch.load(md1_pth))\n",
    "# model2 = LeNet().to(device)\n",
    "# model2.load_state_dict(torch.load(md2_pth))\n",
    "# model3 = LeNet().to(device)\n",
    "# model3.load_state_dict(torch.load(md3_pth))\n",
    "# model4 = LeNet().to(device)\n",
    "# model4.load_state_dict(torch.load(md4_pth))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "66\n",
      "99\n"
     ]
    }
   ],
   "source": [
    "modelpth_set = []\n",
    "round_max = 15\n",
    "client_max = 50\n",
    "for dirpath, dirnames, filenames in os.walk(attack_pth):\n",
    "    # Do something with the files or directories found\n",
    "    for filename in filenames:\n",
    "        # print(os.path.join(dirpath, filename).replace(\"\\\\\",\"/\"))\n",
    "        # if int(re.findall(\"\\d+\\.?\\d*\", filename)[0]) < round_max:\n",
    "        if int(re.findall(r'-?\\d+', filename)[0]) < round_max and int(re.findall(r'-?\\d+', filename)[1]) < client_max:\n",
    "            modelpth_set.append(os.path.join(dirpath, filename).replace(\"\\\\\",\"/\"))\n",
    "attacks = len(modelpth_set)\n",
    "print(len(modelpth_set))\n",
    "for dirpath, dirnames, filenames in os.walk(normal_pth):\n",
    "    # Do something with the files or directories found\n",
    "    for filename in filenames:\n",
    "        # print(os.path.join(dirpath, filename).replace(\"\\\\\",\"/\"))\n",
    "        if int(re.findall(r'-?\\d+', filename)[0]) < round_max and int(re.findall(r'-?\\d+', filename)[1]) < client_max:\n",
    "            modelpth_set.append(os.path.join(dirpath, filename).replace(\"\\\\\",\"/\"))\n",
    "\n",
    "print(len(modelpth_set))\n",
    "normals = len(modelpth_set) - attacks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# print(model1,model3)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# criterion = nn.CrossEntropyLoss()\n",
    "# optimizer = torch.optim.SGD(model1.parameters(), lr=0.01, momentum=0.5)\n",
    "\n",
    "# for X, Y in test_loader:\n",
    "#     X_test = X.to(device)\n",
    "#     Y_test = Y.to(device)\n",
    "#     print(Y)\n",
    "#     break\n",
    "\n",
    "# # Forward pass\n",
    "# outputs = model1(X_test)\n",
    "\n",
    "# # Compute the loss\n",
    "# loss = criterion(outputs, Y_test)\n",
    "\n",
    "# # Zero the gradients\n",
    "# optimizer.zero_grad()\n",
    "\n",
    "# # Backward pass\n",
    "# loss.backward()\n",
    "\n",
    "# # Access the gradients\n",
    "# gradients1 = {}\n",
    "# for name, param in model1.named_parameters():\n",
    "#     gradients1[name] = param.grad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "99\n"
     ]
    }
   ],
   "source": [
    "corr_depth = 1\n",
    "# model = LeNet().to(device)\n",
    "# model.load_state_dict(torch.load(modelpth_set[0]))\n",
    "# for name, w in model.named_parameters():\n",
    "#             if \"weight\" in name:\n",
    "#                 corr_depth += 1\n",
    "\n",
    "corr_max = normals + attacks\n",
    "corr = np.zeros((corr_max, corr_max, corr_depth))\n",
    "\n",
    "gradientlist = []\n",
    "for idx, modelp1 in enumerate(modelpth_set):\n",
    "    gradientlist.append(getGradients(modelp1))\n",
    "\n",
    "print(len(gradientlist))\n",
    "\n",
    "for id1, modelp1 in enumerate(gradientlist):\n",
    "    if id1 >= corr_max:\n",
    "        break\n",
    "    for id2, modelp2 in enumerate(gradientlist):\n",
    "        if id2 >= corr_max:\n",
    "            break\n",
    "        if id2 <= id1:\n",
    "            continue\n",
    "        gradients1 = gradientlist[id1]\n",
    "        gradients2 = gradientlist[id2]\n",
    "        \n",
    "        corrlist = []\n",
    "#         for w in gradients1.keys():\n",
    "#             if \"weight\" in w:\n",
    "#                 acts1 = gradients1[w]\n",
    "#                 acts2 = gradients2[w]\n",
    "#                 # print(\"activation shapes\", acts1.shape, acts2.shape)\n",
    "#                 factors = sympy.factorint(acts1.reshape(-1,1).shape[0])\n",
    "#                 # f = factors.items()[-1]\n",
    "#                 f = int(sorted(factors.keys())[-1]) ** (factors[sorted(factors.keys())[-1]]-1)\n",
    "                \n",
    "#                 acts1 = acts1.cpu().reshape((f,-1))\n",
    "#                 acts2 = acts2.cpu().reshape((f,-1))\n",
    "#                 l = linear_CKA(acts1.T, acts2.T)\n",
    "#                 k = kernel_CKA(acts1.T, acts2.T)\n",
    "#                 # print('Linear CKA: {}'.format(l))\n",
    "#                 # print('RBF Kernel: {}'.format(k))\n",
    "#                 corrlist.append((l+k)/2)\n",
    "        acts1 = gradients1[\"layer3.0.weight\"]\n",
    "        acts2 = gradients2[\"layer3.0.weight\"]\n",
    "        acts1 = acts1.cpu().reshape((50,-1))\n",
    "        acts2 = acts2.cpu().reshape((50,-1))\n",
    "        l = linear_CKA(acts1.T, acts2.T)\n",
    "        k = kernel_CKA(acts1.T, acts2.T)\n",
    "        corrlist.append((l+k)/2)\n",
    "        # print(corrlist)\n",
    "                \n",
    "        corr[id1][id2] = np.array(corrlist)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[[0.        ]\n",
      "  [0.74330697]\n",
      "  [0.8367257 ]\n",
      "  ...\n",
      "  [0.60688541]\n",
      "  [0.75783073]\n",
      "  [0.79232584]]\n",
      "\n",
      " [[0.        ]\n",
      "  [0.        ]\n",
      "  [0.52616814]\n",
      "  ...\n",
      "  [0.49615328]\n",
      "  [0.76272326]\n",
      "  [0.70104906]]\n",
      "\n",
      " [[0.        ]\n",
      "  [0.        ]\n",
      "  [0.        ]\n",
      "  ...\n",
      "  [0.63174296]\n",
      "  [0.61699242]\n",
      "  [0.68088126]]\n",
      "\n",
      " ...\n",
      "\n",
      " [[0.        ]\n",
      "  [0.        ]\n",
      "  [0.        ]\n",
      "  ...\n",
      "  [0.        ]\n",
      "  [0.6675863 ]\n",
      "  [0.79338153]]\n",
      "\n",
      " [[0.        ]\n",
      "  [0.        ]\n",
      "  [0.        ]\n",
      "  ...\n",
      "  [0.        ]\n",
      "  [0.        ]\n",
      "  [0.90227139]]\n",
      "\n",
      " [[0.        ]\n",
      "  [0.        ]\n",
      "  [0.        ]\n",
      "  ...\n",
      "  [0.        ]\n",
      "  [0.        ]\n",
      "  [0.        ]]]\n"
     ]
    }
   ],
   "source": [
    "print(corr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.colorbar.Colorbar at 0x23e51c43850>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(corr, cmap='hot')\n",
    "plt.colorbar()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# criterion = nn.CrossEntropyLoss()\n",
    "# optimizer = torch.optim.SGD(model2.parameters(), lr=0.01, momentum=0.5)\n",
    "\n",
    "# for X, Y in test_loader:\n",
    "#     X_test = X.to(device)\n",
    "#     Y_test = Y.to(device)\n",
    "#     print(Y)\n",
    "#     break\n",
    "\n",
    "# # Forward pass\n",
    "# outputs = model2(X_test)\n",
    "\n",
    "# # Compute the loss\n",
    "# loss = criterion(outputs, Y_test)\n",
    "\n",
    "# # Zero the gradients\n",
    "# optimizer.zero_grad()\n",
    "\n",
    "# # Backward pass\n",
    "# loss.backward()\n",
    "\n",
    "# # Access the gradients\n",
    "# gradients2 = {}\n",
    "# for name, param in model2.named_parameters():\n",
    "#     gradients2[name] = param.grad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# criterion = nn.CrossEntropyLoss()\n",
    "# optimizer = torch.optim.SGD(model3.parameters(), lr=0.01, momentum=0.5)\n",
    "\n",
    "# for X, Y in test_loader:\n",
    "#     X_test = X.to(device)\n",
    "#     Y_test = Y.to(device)\n",
    "#     break\n",
    "\n",
    "# # Forward pass\n",
    "# outputs = model3(X_test)\n",
    "\n",
    "# # Compute the loss\n",
    "# loss = criterion(outputs, Y_test)\n",
    "\n",
    "# # Zero the gradients\n",
    "# optimizer.zero_grad()\n",
    "\n",
    "# # Backward pass\n",
    "# loss.backward()\n",
    "\n",
    "# # Access the gradients\n",
    "# gradients3 = {}\n",
    "# for name, param in model3.named_parameters():\n",
    "#     gradients3[name] = param.grad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "jupyter": {
     "source_hidden": true
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# criterion = nn.CrossEntropyLoss()\n",
    "# optimizer = torch.optim.SGD(model4.parameters(), lr=0.01, momentum=0.5)\n",
    "\n",
    "# for X, Y in test_loader:\n",
    "#     X_test = X.to(device)\n",
    "#     Y_test = Y.to(device)\n",
    "#     break\n",
    "\n",
    "# # Forward pass\n",
    "# outputs = model4(X_test)\n",
    "\n",
    "# # Compute the loss\n",
    "# loss = criterion(outputs, Y_test)\n",
    "\n",
    "# # Zero the gradients\n",
    "# optimizer.zero_grad()\n",
    "\n",
    "# # Backward pass\n",
    "# loss.backward()\n",
    "\n",
    "# # Access the gradients\n",
    "# gradients4 = {}\n",
    "# for name, param in model4.named_parameters():\n",
    "#     gradients4[name] = param.grad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# print(gradients1.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "acts1 = gradients1[\"layer3.0.weight\"]\n",
    "acts2 = gradients2[\"layer3.0.weight\"]\n",
    "acts3 = gradients3[\"layer3.0.weight\"]\n",
    "acts4 = gradients4[\"layer3.0.weight\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"activation shapes\", acts1.shape, acts2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "acts1 = acts1.cpu().reshape((50,-1))\n",
    "acts2 = acts2.cpu().reshape((50,-1))\n",
    "acts3 = acts3.cpu().reshape((50,-1))\n",
    "acts4 = acts4.cpu().reshape((50,-1))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Linear CKA: {}'.format(linear_CKA(acts1.T, acts2.T)))\n",
    "print('RBF Kernel: {}'.format(kernel_CKA(acts1.T, acts2.T)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Linear CKA: {}'.format(linear_CKA(acts4.T, acts3.T)))\n",
    "print('RBF Kernel: {}'.format(kernel_CKA(acts4.T, acts3.T)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Linear CKA: {}'.format(linear_CKA(acts1.T, acts3.T)))\n",
    "print('RBF Kernel: {}'.format(kernel_CKA(acts1.T, acts3.T)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Linear CKA: {}'.format(linear_CKA(acts2.T, acts4.T)))\n",
    "print('Linear CKA: {}'.format(kernel_CKA(acts2.T, acts4.T)))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.13 (default, Mar 28 2022, 11:38:47) \n[GCC 7.5.0]"
  },
  "vscode": {
   "interpreter": {
    "hash": "8bff0657a98b8ee576cbe89028f6a544b770fee234379978e147d89ef60d92ed"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}