M23 Hyperparameter tuning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load all necessary imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-03-26 15:04:41.212758: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1929] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 21897 MB memory:  -> device: 0, name: NVIDIA GeForce RTX 3090, pci bus id: 0000:68:00.0, compute capability: 8.6\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras import layers, models\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "import h5py\n",
    "import wandb\n",
    "import random\n",
    "\n",
    "\n",
    "# Configure Keras to use GPU\n",
    "config = tf.compat.v1.ConfigProto()\n",
    "config.gpu_options.allow_growth = True\n",
    "sess = tf.compat.v1.Session(config=config)\n",
    "tf.compat.v1.keras.backend.set_session(sess)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data Loading and Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Load data\n",
    "with h5py.File('TrainingData5zeroes.h5', 'r') as hdf:\n",
    "    ls = list(hdf.keys())\n",
    "    images = hdf.get('images')\n",
    "    boxes = hdf.get('boxes')\n",
    "    images = np.array(images)\n",
    "    boxes = np.array(boxes)\n",
    "\n",
    "\n",
    "## Normalize images and bounding boxes\n",
    "    \n",
    "image_normalized = (images+1e-9)/9.26 # Normalize images and add noise\n",
    "# Normalize bounding boxes\n",
    "normalized_boxes = boxes / [1,64,64,64,64]\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "images_np = image_normalized\n",
    "#images_np = tf.expand_dims(images_np, axis=-1)\n",
    "probabilities = np.array(normalized_boxes[:, :,:-4])\n",
    "probabilities = tf.expand_dims(probabilities, axis=1)\n",
    "boxes_np = np.array(normalized_boxes[:, :, 1:])\n",
    "boxes_np = tf.expand_dims(boxes_np, axis=1)\n",
    "\n",
    "dataset = tf.data.Dataset.from_tensor_slices((images_np,{'x_prob_reshape':probabilities,'x_boxes_reshape':boxes_np}))\n",
    "dataset = dataset.shuffle(buffer_size=10000,reshuffle_each_iteration=True)\n",
    "dataset = dataset.batch(128)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Model Train\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "input_shape = (64,64,1)\n",
    "num_classes = 280\n",
    "num_coordinates = 4\n",
    "\n",
    "x_input = layers.Input(shape=input_shape)\n",
    "#Layer 1\n",
    "x = layers.Conv2D(64, kernel_size=3, padding='same', activation='relu')(x_input)\n",
    "x = layers.MaxPool2D()(x)\n",
    "x = layers.BatchNormalization()(x) \n",
    "x = layers.Conv2D(64, kernel_size=3, padding='same', activation='relu')(x)\n",
    "#Layer 2\n",
    "x = layers.Conv2D(128, kernel_size=3, padding='same', activation='relu')(x)\n",
    "x = layers.Conv2D(128, kernel_size=3, padding='same', activation='relu')(x)\n",
    "#Layer 3\n",
    "x = layers.Conv2D(256, kernel_size=3, padding='same', activation='relu')(x)\n",
    "x = layers.Conv2D(256, kernel_size=3, padding='same', activation='relu')(x)\n",
    "#Layer 4\n",
    "x = layers.Conv2D(512, kernel_size=3, padding='same', activation='relu')(x)\n",
    "x = layers.MaxPool2D()(x)\n",
    "x = layers.Conv2D(512, kernel_size=3, padding='same', activation='relu')(x)\n",
    "x = layers.MaxPool2D()(x)\n",
    "x = layers.Conv2D(512, kernel_size=3, padding='same', activation='relu')(x)\n",
    "x = layers.MaxPool2D()(x)\n",
    "#Layer 5\n",
    "x = layers.Conv2D(256, kernel_size=5, padding='same', activation='relu')(x)\n",
    "x = layers.MaxPool2D()(x)\n",
    "x = layers.BatchNormalization()(x) \n",
    "\n",
    "\n",
    "x = layers.Flatten()(x)\n",
    "# Probability output\n",
    "x_prob = layers.Dense(num_classes, activation='sigmoid', name='x_prob')(x)\n",
    "x_prob_reshape = layers.Reshape((-1, num_classes, 1), name='x_prob_reshape')(x_prob)\n",
    "\n",
    "# Bounding box output\n",
    "x_boxes = layers.Dense(num_classes * num_coordinates, activation='sigmoid', name='x_boxes')(x)\n",
    "x_boxes_reshape = layers.Reshape((-1, num_classes, num_coordinates), name='x_boxes_reshape')(x_boxes)\n",
    "\n",
    "\n",
    "model = tf.keras.models.Model(x_input, [x_prob_reshape, x_boxes_reshape])\n",
    "optimizer = tf.keras.optimizers.Adam(learning_rate=3e-5) \n",
    "model.compile(optimizer= optimizer, loss= {'x_prob_reshape': tf.keras.losses.BinaryCrossentropy(), 'x_boxes_reshape':tf.keras.losses.MeanSquaredError()}, metrics={'x_prob_reshape': 'accuracy'} )\n",
    "num_epochs = 20\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "sweep_config = {\n",
    "    'method': 'random'\n",
    "    }\n",
    "metric = {\n",
    "    'name': 'loss',\n",
    "    'goal': 'minimize'   \n",
    "    }\n",
    "\n",
    "sweep_config['metric'] = metric\n",
    "parameters_dict = {\n",
    "    'optimizer': {\n",
    "        'values': ['adam', 'sgd']\n",
    "        },\n",
    "    'fc_layer_size': {\n",
    "        'values': [128, 256, 512]\n",
    "        },\n",
    "    'dropout': {\n",
    "          'values': [0.3, 0.4, 0.5]\n",
    "        },\n",
    "    }\n",
    "\n",
    "sweep_config['parameters'] = parameters_dict\n",
    "parameters_dict.update({\n",
    "    'epochs': {\n",
    "        'value': 1}\n",
    "    })\n",
    "parameters_dict.update({\n",
    "    'learning_rate': {\n",
    "        # a flat distribution between 0 and 0.1\n",
    "        'distribution': 'uniform',\n",
    "        'min': 0,\n",
    "        'max': 0.1\n",
    "      },\n",
    "    'batch_size': {\n",
    "        # integers between 32 and 256\n",
    "        # with evenly-distributed logarithms \n",
    "        'distribution': 'q_log_uniform_values',\n",
    "        'q': 8,\n",
    "        'min': 32,\n",
    "        'max': 256,\n",
    "      }\n",
    "    })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'method': 'random',\n",
      " 'metric': {'goal': 'minimize', 'name': 'loss'},\n",
      " 'parameters': {'batch_size': {'distribution': 'q_log_uniform_values',\n",
      "                               'max': 256,\n",
      "                               'min': 32,\n",
      "                               'q': 8},\n",
      "                'dropout': {'values': [0.3, 0.4, 0.5]},\n",
      "                'epochs': {'value': 1},\n",
      "                'fc_layer_size': {'values': [128, 256, 512]},\n",
      "                'learning_rate': {'distribution': 'uniform',\n",
      "                                  'max': 0.1,\n",
      "                                  'min': 0},\n",
      "                'optimizer': {'values': ['adam', 'sgd']}}}\n"
     ]
    }
   ],
   "source": [
    "import pprint\n",
    "\n",
    "pprint.pprint(sweep_config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Create sweep with ID: g9lv4jfu\n",
      "Sweep URL: https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu\n"
     ]
    }
   ],
   "source": [
    "sweep_id = wandb.sweep(sweep_config, project=\"M23 Hyperparameter Tuning\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# wandb.init(\n",
    "#     # set the wandb project where this run will be logged\n",
    "#     project=\"MyFirstComputerVision\",\n",
    "    \n",
    "#     # track hyperparameters and run metadata\n",
    "#     config={\n",
    "#     \"learning_rate\": 3e-5,\n",
    "#     \"architecture\": \"model\",\n",
    "#     \"dataset\": \"dataset\",\n",
    "#     \"epochs\": \"num_epochs\",\n",
    "#     }\n",
    "# )\n",
    "# # simulate training\n",
    "\n",
    "# offset = random.random() / 5\n",
    "# for epoch in range(2,num_epochs):\n",
    "#     acc = 1 - 2 ** -epoch - random.random() / epoch - offset\n",
    "#     loss = 2 ** -epoch + random.random() / epoch + offset\n",
    "    \n",
    "#     # log metrics to wandb\n",
    "#     wandb.log({\"acc\": acc, \"loss\": loss})\n",
    "    \n",
    "# # [optional] finish the wandb run, necessary in notebooks\n",
    "# wandb.finish()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.models.load_model('M11overfittedmodel3variant.h5')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-03-26 13:50:47.516518: I external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:454] Loaded cuDNN version 8904\n",
      "2024-03-26 13:50:47.602849: I external/local_tsl/tsl/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory\n",
      "2024-03-26 13:50:48.622823: I external/local_tsl/tsl/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory\n",
      "2024-03-26 13:50:49.512318: I external/local_xla/xla/service/service.cc:168] XLA service 0x7cecd44557e0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
      "2024-03-26 13:50:49.512356: I external/local_xla/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3090, Compute Capability 8.6\n",
      "2024-03-26 13:50:49.525351: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:269] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
      "WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\n",
      "I0000 00:00:1711475449.672995   55592 device_compiler.h:186] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "79/79 [==============================] - 21s 93ms/step - loss: 0.5040 - x_prob_reshape_loss: 0.3691 - x_boxes_reshape_loss: 0.1349 - x_prob_reshape_accuracy: 0.8537\n",
      "Epoch 2/10\n",
      "79/79 [==============================] - 6s 71ms/step - loss: 0.2388 - x_prob_reshape_loss: 0.1736 - x_boxes_reshape_loss: 0.0651 - x_prob_reshape_accuracy: 0.9482\n",
      "Epoch 3/10\n",
      "79/79 [==============================] - 6s 72ms/step - loss: 0.1882 - x_prob_reshape_loss: 0.1345 - x_boxes_reshape_loss: 0.0537 - x_prob_reshape_accuracy: 0.9566\n",
      "Epoch 4/10\n",
      "79/79 [==============================] - 6s 71ms/step - loss: 0.1692 - x_prob_reshape_loss: 0.1191 - x_boxes_reshape_loss: 0.0501 - x_prob_reshape_accuracy: 0.9581\n",
      "Epoch 5/10\n",
      "79/79 [==============================] - 6s 70ms/step - loss: 0.1578 - x_prob_reshape_loss: 0.1096 - x_boxes_reshape_loss: 0.0483 - x_prob_reshape_accuracy: 0.9592\n",
      "Epoch 6/10\n",
      "79/79 [==============================] - 6s 71ms/step - loss: 0.1443 - x_prob_reshape_loss: 0.0975 - x_boxes_reshape_loss: 0.0468 - x_prob_reshape_accuracy: 0.9643\n",
      "Epoch 7/10\n",
      "79/79 [==============================] - 6s 72ms/step - loss: 0.1393 - x_prob_reshape_loss: 0.0932 - x_boxes_reshape_loss: 0.0460 - x_prob_reshape_accuracy: 0.9646\n",
      "Epoch 8/10\n",
      "79/79 [==============================] - 6s 72ms/step - loss: 0.1319 - x_prob_reshape_loss: 0.0867 - x_boxes_reshape_loss: 0.0452 - x_prob_reshape_accuracy: 0.9667\n",
      "Epoch 9/10\n",
      "79/79 [==============================] - 6s 73ms/step - loss: 0.1283 - x_prob_reshape_loss: 0.0836 - x_boxes_reshape_loss: 0.0447 - x_prob_reshape_accuracy: 0.9674\n",
      "Epoch 10/10\n",
      "79/79 [==============================] - 6s 73ms/step - loss: 0.1240 - x_prob_reshape_loss: 0.0798 - x_boxes_reshape_loss: 0.0442 - x_prob_reshape_accuracy: 0.9689\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "\n",
    "# model.fit(\n",
    "#     dataset,\n",
    "#     epochs=num_epochs \n",
    "   \n",
    "# )\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[34m\u001b[1mwandb\u001b[0m: Agent Starting Run: e4ltebwn with config:\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tbatch_size: 48\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tdropout: 0.5\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tepochs: 1\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tfc_layer_size: 512\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tlearning_rate: 0.03896386007651949\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \toptimizer: adam\n",
      "Failed to detect the name of this notebook, you can set it manually with the WANDB_NOTEBOOK_NAME environment variable to enable code saving.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "Tracking run with wandb version 0.16.5"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Run data is saved locally in <code>/home/m3-learning/wandb/run-20240326_161146-e4ltebwn</code>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Syncing run <strong><a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/e4ltebwn/workspace' target=\"_blank\">bright-sweep-10</a></strong> to <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>Sweep page: <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View project at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View sweep at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View run at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/e4ltebwn/workspace' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/e4ltebwn/workspace</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Traceback (most recent call last):\n",
      "  File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "    model = model(config)\n",
      "UnboundLocalError: local variable 'model' referenced before assignment\n"
     ]
    },
    {
     "data": {
      "text/html": [
       " View run <strong style=\"color:#cdcd00\">bright-sweep-10</strong> at: <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/e4ltebwn/workspace' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/e4ltebwn/workspace</a><br/>Synced 6 W&B file(s), 0 media file(s), 0 artifact file(s) and 0 other file(s)"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Find logs at: <code>./wandb/run-20240326_161146-e4ltebwn/logs</code>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Run e4ltebwn errored:\n",
      "Traceback (most recent call last):\n",
      "  File \"/home/m3-learning/anaconda3/envs/tiny_yolo/lib/python3.10/site-packages/wandb/agents/pyagent.py\", line 308, in _run_job\n",
      "    self._function()\n",
      "  File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "    model = model(config)\n",
      "UnboundLocalError: local variable 'model' referenced before assignment\n",
      "\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m Run e4ltebwn errored:\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m Traceback (most recent call last):\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m   File \"/home/m3-learning/anaconda3/envs/tiny_yolo/lib/python3.10/site-packages/wandb/agents/pyagent.py\", line 308, in _run_job\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m     self._function()\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m   File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m     model = model(config)\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m UnboundLocalError: local variable 'model' referenced before assignment\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m \n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: Agent Starting Run: a8kic56w with config:\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tbatch_size: 40\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tdropout: 0.5\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tepochs: 1\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tfc_layer_size: 512\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tlearning_rate: 0.04961518314473029\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \toptimizer: adam\n",
      "Failed to detect the name of this notebook, you can set it manually with the WANDB_NOTEBOOK_NAME environment variable to enable code saving.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "Tracking run with wandb version 0.16.5"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Run data is saved locally in <code>/home/m3-learning/wandb/run-20240326_161159-a8kic56w</code>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Syncing run <strong><a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/a8kic56w/workspace' target=\"_blank\">usual-sweep-11</a></strong> to <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>Sweep page: <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View project at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View sweep at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View run at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/a8kic56w/workspace' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/a8kic56w/workspace</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Traceback (most recent call last):\n",
      "  File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "    model = model(config)\n",
      "UnboundLocalError: local variable 'model' referenced before assignment\n"
     ]
    },
    {
     "data": {
      "text/html": [
       " View run <strong style=\"color:#cdcd00\">usual-sweep-11</strong> at: <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/a8kic56w/workspace' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/a8kic56w/workspace</a><br/>Synced 6 W&B file(s), 0 media file(s), 0 artifact file(s) and 0 other file(s)"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Find logs at: <code>./wandb/run-20240326_161159-a8kic56w/logs</code>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Run a8kic56w errored:\n",
      "Traceback (most recent call last):\n",
      "  File \"/home/m3-learning/anaconda3/envs/tiny_yolo/lib/python3.10/site-packages/wandb/agents/pyagent.py\", line 308, in _run_job\n",
      "    self._function()\n",
      "  File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "    model = model(config)\n",
      "UnboundLocalError: local variable 'model' referenced before assignment\n",
      "\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m Run a8kic56w errored:\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m Traceback (most recent call last):\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m   File \"/home/m3-learning/anaconda3/envs/tiny_yolo/lib/python3.10/site-packages/wandb/agents/pyagent.py\", line 308, in _run_job\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m     self._function()\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m   File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m     model = model(config)\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m UnboundLocalError: local variable 'model' referenced before assignment\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m \n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: Agent Starting Run: gtinsi5u with config:\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tbatch_size: 96\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tdropout: 0.4\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tepochs: 1\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tfc_layer_size: 256\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \tlearning_rate: 0.07269372747771771\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \toptimizer: adam\n",
      "Failed to detect the name of this notebook, you can set it manually with the WANDB_NOTEBOOK_NAME environment variable to enable code saving.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "Tracking run with wandb version 0.16.5"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Run data is saved locally in <code>/home/m3-learning/wandb/run-20240326_161211-gtinsi5u</code>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Syncing run <strong><a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/gtinsi5u/workspace' target=\"_blank\">fancy-sweep-12</a></strong> to <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>Sweep page: <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View project at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View sweep at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/sweeps/g9lv4jfu</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       " View run at <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/gtinsi5u/workspace' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/gtinsi5u/workspace</a>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Traceback (most recent call last):\n",
      "  File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "    model = model(config)\n",
      "UnboundLocalError: local variable 'model' referenced before assignment\n"
     ]
    },
    {
     "data": {
      "text/html": [
       " View run <strong style=\"color:#cdcd00\">fancy-sweep-12</strong> at: <a href='https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/gtinsi5u/workspace' target=\"_blank\">https://wandb.ai/alphanium/M23%20Hyperparameter%20Tuning/runs/gtinsi5u/workspace</a><br/>Synced 6 W&B file(s), 0 media file(s), 0 artifact file(s) and 0 other file(s)"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "Find logs at: <code>./wandb/run-20240326_161211-gtinsi5u/logs</code>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Run gtinsi5u errored:\n",
      "Traceback (most recent call last):\n",
      "  File \"/home/m3-learning/anaconda3/envs/tiny_yolo/lib/python3.10/site-packages/wandb/agents/pyagent.py\", line 308, in _run_job\n",
      "    self._function()\n",
      "  File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "    model = model(config)\n",
      "UnboundLocalError: local variable 'model' referenced before assignment\n",
      "\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m Run gtinsi5u errored:\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m Traceback (most recent call last):\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m   File \"/home/m3-learning/anaconda3/envs/tiny_yolo/lib/python3.10/site-packages/wandb/agents/pyagent.py\", line 308, in _run_job\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m     self._function()\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m   File \"/tmp/ipykernel_59888/1486553214.py\", line 6, in train\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m     model = model(config)\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m UnboundLocalError: local variable 'model' referenced before assignment\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m \n",
      "Detected 3 failed runs in the first 60 seconds, killing sweep.\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[32m\u001b[41mERROR\u001b[0m Detected 3 failed runs in the first 60 seconds, killing sweep.\n",
      "\u001b[34m\u001b[1mwandb\u001b[0m: To disable this check set WANDB_AGENT_DISABLE_FLAPPING=true\n"
     ]
    }
   ],
   "source": [
    "def train():\n",
    "    # Initialize a new wandb run\n",
    "    with wandb.init() as run:\n",
    "        config = run.config\n",
    "        # Build, train, and evaluate your model\n",
    "        model = model(config)\n",
    "        model.fit(dataset, epochs=config.num_epochs, batch_size=config.batch_size)\n",
    "\n",
    "wandb.agent(sweep_id, train, count=5)  # Run the sweep agent\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model.save('M11overfittedmodel3variant.h5')\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model.save('M11overfittedmodel3variant.keras')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10000, 280, 5)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "boxes.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "313/313 [==============================] - 3s 8ms/step\n"
     ]
    }
   ],
   "source": [
    "dataset = tf.data.Dataset.from_tensor_slices((images_np,{'x_prob_reshape':probabilities,'x_boxes_reshape':boxes_np}))\n",
    "\n",
    "# # Assuming dataset has 10,000 examples and we want an 80/20 split\n",
    "# total_items = 10000\n",
    "# train_size = int(total_items * 0.6)\n",
    "# # val_size = total_items - train_size\n",
    "\n",
    "# # Splitting the dataset\n",
    "# train_dataset = dataset.take(train_size)\n",
    "# # val_dataset = dataset.skip(train_size)\n",
    "\n",
    "\n",
    "# # train_dataset = train_dataset.shuffle(buffer_size=train_size,reshuffle_each_iteration=True)\n",
    "# train_dataset = train_dataset.batch(6000)\n",
    "# inputs,targets = next(iter(train_dataset))\n",
    "dataset = dataset.batch(10000)\n",
    "inputs,targets = next(iter(dataset))\n",
    "output = model.predict(inputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: shape=(1, 280, 4), dtype=float64, numpy=\n",
       "array([[[0.734375, 0.015625, 0.734375, 0.03125 ],\n",
       "        [0.      , 0.      , 0.      , 0.      ],\n",
       "        [0.      , 0.      , 0.      , 0.      ],\n",
       "        ...,\n",
       "        [0.      , 0.      , 0.      , 0.      ],\n",
       "        [0.      , 0.      , 0.      , 0.      ],\n",
       "        [0.      , 0.      , 0.      , 0.      ]]])>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "targets['x_boxes_reshape'][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[7.24306941e-01, 3.01006213e-02, 7.41804957e-01, 3.33228447e-02],\n",
       "        [2.24880949e-02, 7.51386629e-03, 2.32093632e-02, 7.64035154e-03],\n",
       "        [4.27702442e-03, 6.26132824e-03, 4.65806713e-03, 6.90292520e-03],\n",
       "        ...,\n",
       "        [2.88477138e-04, 1.39197495e-04, 1.47837753e-04, 1.58194220e-04],\n",
       "        [1.42245786e-04, 2.08800062e-04, 1.84095348e-04, 1.41638870e-04],\n",
       "        [1.16996292e-04, 1.13774906e-04, 1.76820322e-04, 1.51523607e-04]]],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "output[1][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MSE Loss: 1.3558886642682864e-07\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "\n",
    "r = np.random.randint(0,100)\n",
    "tensor1 = tf.constant(targets['x_prob_reshape'], dtype=tf.float64)\n",
    "\n",
    "tensor2 = tf.constant(output[0], dtype=tf.float32)\n",
    "\n",
    "\n",
    "tensor2 = tf.cast(tensor2, tf.float64)\n",
    "\n",
    "\n",
    "mse_loss_fn =  tf.keras.losses.BinaryCrossentropy()\n",
    "mse_loss = mse_loss_fn(tensor1, tensor2)\n",
    "\n",
    "print(\"MSE Loss:\", mse_loss.numpy())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 400x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 400x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def visualize_bounding_boxes(image, probability_vector, bounding_box_coordinates, threshold=0.9):\n",
    "    \"\"\"\n",
    "    Visualizes bounding boxes on an image based on a probability vector.\n",
    "\n",
    "    Parameters:\n",
    "    - image: A 3D tensor representing the image.\n",
    "    - probability_vector: A 1D tensor representing the probabilities associated with bounding boxes.\n",
    "    - bounding_box_coordinates: A 2D tensor representing bounding box coordinates.\n",
    "    - threshold: Probability threshold for visualization.\n",
    "\n",
    "    Returns:\n",
    "    None (displays the image with bounding boxes).\n",
    "    \"\"\"\n",
    "    # Convert to NumPy arrays for easier handling\n",
    "    image_np = image\n",
    "    prob_vector_np = probability_vector\n",
    "    bbox_coordinates_np = bounding_box_coordinates\n",
    "   # Denormalize image if necessary (adjust based on your normalization method)\n",
    "    denormalized_image = image_np  # Modify if normalization was applied during training\n",
    "\n",
    "    # Visualize the image\n",
    "    plt.figure(figsize=(4, 4))\n",
    "    plt.imshow(denormalized_image, cmap='gray')\n",
    "    plt.title(\"Bounding Box Visualization\")\n",
    "\n",
    "    # Plot bounding boxes based on probability threshold\n",
    "    for i in range(len(prob_vector_np)):\n",
    "        prob = prob_vector_np[i]\n",
    "        bbox = bbox_coordinates_np[i]\n",
    "        if prob > threshold:\n",
    "            # Denormalize bounding box coordinates if necessary\n",
    "            denormalized_bbox = bbox  # Modify if normalization was applied during training\n",
    "            y1, x1, y2, x2 = denormalized_bbox\n",
    "            plt.plot([x1, x2, x2, x1, x1],[y1, y1, y2, y2, y1],\n",
    "                     \n",
    "                     color='r', linewidth=2, label='Bounding Box')\n",
    "\n",
    "   \n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "   \n",
    "    plt.show()\n",
    "\n",
    "t = np.random.randint(0,100)\n",
    "\n",
    "\n",
    "visualize_bounding_boxes(tf.convert_to_tensor(inputs[t]), probabilities.numpy()[t].squeeze(), tf.convert_to_tensor(output[1][t,0,:,:])*[64,64,64,64]) ##myprediction\n",
    "visualize_bounding_boxes(tf.convert_to_tensor(image_normalized[t]), probabilities.numpy()[t].squeeze(), tf.convert_to_tensor(boxes_np[t,0,:,:])*[64,64,64,64]) ##myprediction\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tiny_yolo",
   "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": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}