notebooks/simple_demo.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "from torch.utils.data import DataLoader\n",
    "\n",
    "from cupbearer import data, detectors, models, scripts, tasks, utils"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training a backdoored classifier\n",
    "First, we train a classifier on poisoned data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = data.MNIST()\n",
    "val_data = data.MNIST(train=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = models.MLP(input_shape=(28, 28), hidden_dims=[128, 128], output_dim=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "GPU available: True (mps), used: True\n",
      "TPU available: False, using: 0 TPU cores\n",
      "IPU available: False, using: 0 IPUs\n",
      "HPU available: False, using: 0 HPUs\n",
      "\n",
      "  | Name           | Type               | Params\n",
      "------------------------------------------------------\n",
      "0 | model          | MLP                | 118 K \n",
      "1 | train_accuracy | MulticlassAccuracy | 0     \n",
      "2 | val_accuracy   | ModuleList         | 0     \n",
      "3 | test_accuracy  | ModuleList         | 0     \n",
      "------------------------------------------------------\n",
      "118 K     Trainable params\n",
      "0         Non-trainable params\n",
      "118 K     Total params\n",
      "0.473     Total estimated model params size (MB)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f3fbd649f84545b79393518b1875ea71",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Sanity Checking: |          | 0/? [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/erik/.pyenv/versions/3.10.9/envs/cupbearer/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:441: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.\n",
      "/Users/erik/.pyenv/versions/3.10.9/envs/cupbearer/lib/python3.10/site-packages/torchmetrics/functional/classification/accuracy.py:77: UserWarning: MPS: no support for int64 reduction ops, casting it to int32 (Triggered internally at /Users/runner/work/pytorch/pytorch/pytorch/aten/src/ATen/native/mps/operations/ReduceOps.mm:144.)\n",
      "  tp = tp.sum(dim=0 if multidim_average == \"global\" else 1)\n",
      "/Users/erik/.pyenv/versions/3.10.9/envs/cupbearer/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:441: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "94e198d2efdb4899a5cc3829dc6a90b7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training: |          | 0/? [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6218ec07439243379c33377a00ade093",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: |          | 0/? [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c1d3eeecce294a499dca8d392f36b779",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: |          | 0/? [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bc1bc65539a04d57ab5b36746317d188",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: |          | 0/? [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "`Trainer.fit` stopped: `max_epochs=3` reached.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'train/loss': tensor(0.0756),\n",
       " 'train/acc_step': tensor(0.9688),\n",
       " 'clean/loss/dataloader_idx_0': tensor(0.1696),\n",
       " 'clean/acc_step/dataloader_idx_0': tensor(0.9468),\n",
       " 'backdoor/loss/dataloader_idx_1': tensor(0.1274),\n",
       " 'backdoor/acc_step/dataloader_idx_1': tensor(0.9580),\n",
       " 'clean/acc_epoch': tensor(0.9468),\n",
       " 'backdoor/acc_epoch': tensor(0.9580),\n",
       " 'train/acc_epoch': tensor(0.9417)}"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scripts.train_classifier(\n",
    "    path=(classifier_path := utils.log_path(\"logs/demo/classifier\")),\n",
    "    model=model,\n",
    "    train_loader=DataLoader(\n",
    "        data.BackdoorDataset(\n",
    "            # Poison 5% of the training data\n",
    "            original=train_data,\n",
    "            backdoor=data.CornerPixelBackdoor(p_backdoor=0.05),\n",
    "        ),\n",
    "        batch_size=64,\n",
    "        shuffle=True,\n",
    "    ),\n",
    "    num_classes=10,\n",
    "    val_loaders={\n",
    "        \"clean\": DataLoader(val_data, batch_size=1024, shuffle=False),\n",
    "        \"backdoor\": DataLoader(\n",
    "            data.BackdoorDataset(\n",
    "                # By default, the poison rate is 100%, so this will let us evaluate\n",
    "                # performance on completely poisoned data\n",
    "                original=val_data,\n",
    "                backdoor=data.CornerPixelBackdoor(),\n",
    "            ),\n",
    "            batch_size=1024,\n",
    "            shuffle=False,\n",
    "        ),\n",
    "    },\n",
    "    max_epochs=3,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can also explicitly evaluate the trained model (right now this is pretty limited and doesn't support multiple datasets at once). In this case it doesn't tell us anything new, but it can be useful if we want to evaluate a model on additional data later:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "GPU available: True (mps), used: True\n",
      "TPU available: False, using: 0 TPU cores\n",
      "IPU available: False, using: 0 IPUs\n",
      "HPU available: False, using: 0 HPUs\n",
      "/Users/erik/.pyenv/versions/3.10.9/envs/cupbearer/lib/python3.10/site-packages/lightning/pytorch/trainer/connectors/data_connector.py:441: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "58498c048eec46bc8fc53c271454eaf3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Testing: |          | 0/? [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓\n",
       "┃<span style=\"font-weight: bold\">        Test metric        </span>┃<span style=\"font-weight: bold\">       DataLoader 0        </span>┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩\n",
       "│<span style=\"color: #008080; text-decoration-color: #008080\">      test/acc_epoch       </span>│<span style=\"color: #800080; text-decoration-color: #800080\">    0.9467999935150146     </span>│\n",
       "│<span style=\"color: #008080; text-decoration-color: #008080\">       test/acc_step       </span>│<span style=\"color: #800080; text-decoration-color: #800080\">    0.9467999935150146     </span>│\n",
       "│<span style=\"color: #008080; text-decoration-color: #008080\">         test/loss         </span>│<span style=\"color: #800080; text-decoration-color: #800080\">    0.16958695650100708    </span>│\n",
       "└───────────────────────────┴───────────────────────────┘\n",
       "</pre>\n"
      ],
      "text/plain": [
       "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓\n",
       "┃\u001b[1m \u001b[0m\u001b[1m       Test metric       \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m      DataLoader 0       \u001b[0m\u001b[1m \u001b[0m┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩\n",
       "│\u001b[36m \u001b[0m\u001b[36m     test/acc_epoch      \u001b[0m\u001b[36m \u001b[0m│\u001b[35m \u001b[0m\u001b[35m   0.9467999935150146    \u001b[0m\u001b[35m \u001b[0m│\n",
       "│\u001b[36m \u001b[0m\u001b[36m      test/acc_step      \u001b[0m\u001b[36m \u001b[0m│\u001b[35m \u001b[0m\u001b[35m   0.9467999935150146    \u001b[0m\u001b[35m \u001b[0m│\n",
       "│\u001b[36m \u001b[0m\u001b[36m        test/loss        \u001b[0m\u001b[36m \u001b[0m│\u001b[35m \u001b[0m\u001b[35m   0.16958695650100708   \u001b[0m\u001b[35m \u001b[0m│\n",
       "└───────────────────────────┴───────────────────────────┘\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "scripts.eval_classifier(\n",
    "    data=val_data,\n",
    "    model=model,\n",
    "    path=classifier_path,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "These results will also have been stored to `<log path>/eval.json` if we want to process them further (e.g. to compare many runs):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[{'test/loss': 0.16958695650100708, 'test/acc_step': 0.9467999935150146, 'test/acc_epoch': 0.9467999935150146}]\n"
     ]
    }
   ],
   "source": [
    "with open(classifier_path / \"eval.json\") as f:\n",
    "    print(json.load(f))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training a backdoor detector\n",
    "We'll train a very simple detector using the Mahalanobis distance. Our model is still in memory, but just for demonstration let's load it again:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initialize a new model with the same architecture as before:\n",
    "model = models.MLP(input_shape=(28, 28), hidden_dims=[128, 128], output_dim=10)\n",
    "# Load the weights:\n",
    "models.load(model, classifier_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 59/59 [00:06<00:00,  9.09it/s]\n",
      "\u001b[32m2024-03-02 18:08:41.589\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mcupbearer.detectors.anomaly_detector\u001b[0m:\u001b[36msave_weights\u001b[0m:\u001b[36m220\u001b[0m - \u001b[1mSaving detector to logs/demo/detector/2024-03-02_18-08-34/detector\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ec6c3dcf6ee742048eeb88a37650d650",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Evaluating:   0%|          | 0/10 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2024-03-02 18:08:41.947\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mcupbearer.detectors.anomaly_detector\u001b[0m:\u001b[36meval\u001b[0m:\u001b[36m139\u001b[0m - \u001b[1mAUC_ROC: 1.0000\u001b[0m\n",
      "\u001b[32m2024-03-02 18:08:41.949\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mcupbearer.detectors.anomaly_detector\u001b[0m:\u001b[36meval\u001b[0m:\u001b[36m140\u001b[0m - \u001b[1mAP: 1.0000\u001b[0m\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "scripts.train_detector(\n",
    "    save_path=(detector_path := utils.log_path(\"logs/demo/detector\")),\n",
    "    task=tasks.backdoor_detection(\n",
    "        model, train_data, val_data, data.CornerPixelBackdoor()\n",
    "    ),\n",
    "    detector=detectors.MahalanobisDetector(),\n",
    "    num_classes=10,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we can see, this was a trivial detection task. As an ablation, we can test whether the detector specifically flags backdoored inputs as anomalous, or just anything out of distribution. Let's again reload the detector just to show how that works:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2024-03-02 18:08:42.519\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mcupbearer.detectors.anomaly_detector\u001b[0m:\u001b[36mload_weights\u001b[0m:\u001b[36m224\u001b[0m - \u001b[1mLoading detector from logs/demo/detector/2024-03-02_18-08-34/detector\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "detector = detectors.MahalanobisDetector()\n",
    "# TODO: The fact that weights are saved in \"detector\" is just a convention used by\n",
    "# the train_detector script, this is kind of weird.\n",
    "detector.load_weights(detector_path / \"detector\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2024-03-02 18:08:43.409\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mcupbearer.detectors.anomaly_detector\u001b[0m:\u001b[36meval\u001b[0m:\u001b[36m139\u001b[0m - \u001b[1mAUC_ROC: 0.9940\u001b[0m\n",
      "\u001b[32m2024-03-02 18:08:43.409\u001b[0m | \u001b[1mINFO    \u001b[0m | \u001b[36mcupbearer.detectors.anomaly_detector\u001b[0m:\u001b[36meval\u001b[0m:\u001b[36m140\u001b[0m - \u001b[1mAP: 0.9784\u001b[0m\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "scripts.eval_detector(\n",
    "    detector=detector,\n",
    "    # We save to a different directory to avoid overwriting the existing default eval:\n",
    "    save_path=detector_path / \"ood_eval\",\n",
    "    task=tasks.Task.from_separate_data(\n",
    "        model=model,\n",
    "        # TODO: this won't actually be used, plausibly Tasks should be split better\n",
    "        # into their training and test data.\n",
    "        trusted_data=train_data,\n",
    "        # Our anomalous data is the backdoor data from above, except we use the\n",
    "        # MNIST test split.\n",
    "        anomalous_test_data=data.BackdoorDataset(\n",
    "            original=val_data,\n",
    "            backdoor=data.CornerPixelBackdoor(),\n",
    "        ),\n",
    "        # Our normal data is MNIST with added noise, this makes the images OOD\n",
    "        # but they shouldn't be mechanistically anomalous.\n",
    "        clean_test_data=data.TransformDataset(val_data, data.GaussianNoise(0.3)),\n",
    "    ),\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we can see, adding noise did make the images quite a bit more \"anomalous\" according to our detector (the blue histogram has shifted to the right to higher anomaly scores). But we still have a very clear separation between these \"merely noisy\" inputs and the backdoored inputs. (This is a very easy to detect backdoor.)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "cupbearer",
   "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": "3.10.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}