{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "7dc067e7",
   "metadata": {},
   "source": [
    "## Gradio Pets"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "05972144",
   "metadata": {},
   "source": [
    "Since the `learn.fine_tune(3)` took very long time in local machine, running it in Kaggle.\n",
    "\n",
    "- Hugging Face - Org Source: https://huggingface.co/spaces/prasanthntu/pets/blob/main/train.ipynb\n",
    "- Kaggle link: https://www.kaggle.com/prasanth07/fastai-prd-apps-pets-training\n",
    "\n",
    "Part of `fastai-course22` > Lesson 3: https://github.com/prasanth-ntu/fastai-course22/blob/master/kaggle_version/03-which-image-models-are-best.ipynb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "11b5a5c3",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-12-09T14:15:12.722396Z",
     "start_time": "2023-12-09T14:15:10.027227Z"
    },
    "execution": {
     "iopub.execute_input": "2023-06-10T08:54:28.429796Z",
     "iopub.status.busy": "2023-06-10T08:54:28.428930Z",
     "iopub.status.idle": "2023-06-10T08:54:35.857170Z",
     "shell.execute_reply": "2023-06-10T08:54:35.856275Z",
     "shell.execute_reply.started": "2023-06-10T08:54:28.429766Z"
    }
   },
   "outputs": [],
   "source": [
    "from fastai.vision.all import *\n",
    "import timm\n",
    "import os, sys\n",
    "import numpy as np\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "28b2fa9d",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-12-09T14:15:13.046568Z",
     "start_time": "2023-12-09T14:15:12.724748Z"
    },
    "execution": {
     "iopub.execute_input": "2023-06-10T08:54:35.865899Z",
     "iopub.status.busy": "2023-06-10T08:54:35.862479Z",
     "iopub.status.idle": "2023-06-10T08:55:08.285776Z",
     "shell.execute_reply": "2023-06-10T08:55:08.284743Z",
     "shell.execute_reply.started": "2023-06-10T08:54:35.865862Z"
    }
   },
   "outputs": [],
   "source": [
    "path = untar_data(URLs.PETS)/'images'\n",
    "\n",
    "dls = ImageDataLoaders.from_name_func('.',\n",
    "    get_image_files(path), valid_pct=0.2, seed=42,\n",
    "    # Regex captures the part of the filename before the underscore followed by decimal values, \n",
    "    # which typically represents the class of the image (e.g., basset_hound_112.jpg > basset_bound)\n",
    "    label_func=RegexLabeller(pat = r'^([^/]+)_\\d+'), \n",
    "    item_tfms=Resize(224))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "472c5e8d",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-12-09T14:19:19.309791Z",
     "start_time": "2023-12-09T14:19:19.295796Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "path: /Users/prasanth.thangavel/.fastai/data/oxford-iiit-pet/images\n",
      "sample files: ['Egyptian_Mau_167.jpg', 'pug_52.jpg', 'basset_hound_112.jpg', 'Siamese_193.jpg', 'shiba_inu_122.jpg']\n",
      "No. of unique pet types: 35\n",
      "Sample pet types: ['Abyssinian', 'Bengal', 'Birman', 'Bombay', 'British']\n"
     ]
    }
   ],
   "source": [
    "print(f\"path: {path}\")\n",
    "\n",
    "# List the contents of the directory\n",
    "files = os.listdir(path)\n",
    "print (f\"sample files: {files[:5]}\")\n",
    "labels = [file.split('_')[0] for file in files]\n",
    "labels_unique = np.unique(labels)\n",
    "print (f\"No. of unique pet types: {len(labels_unique)}\")\n",
    "print (f\"Sample pet types: {list(labels_unique[:5])}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fefc6186",
   "metadata": {
    "ExecuteTime": {
     "start_time": "2023-12-09T14:19:23.273Z"
    },
    "execution": {
     "iopub.execute_input": "2023-06-10T08:55:08.293077Z",
     "iopub.status.busy": "2023-06-10T08:55:08.290699Z",
     "iopub.status.idle": "2023-06-10T08:55:10.941056Z",
     "shell.execute_reply": "2023-06-10T08:55:10.940257Z",
     "shell.execute_reply.started": "2023-06-10T08:55:08.293042Z"
    }
   },
   "outputs": [],
   "source": [
    "dls.show_batch(max_n=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "72c0dfe0",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-06-10T07:08:36.740818Z",
     "start_time": "2023-06-10T07:02:49.332510Z"
    },
    "execution": {
     "iopub.execute_input": "2023-06-10T08:55:10.943412Z",
     "iopub.status.busy": "2023-06-10T08:55:10.942959Z",
     "iopub.status.idle": "2023-06-10T08:59:37.724361Z",
     "shell.execute_reply": "2023-06-10T08:59:37.723260Z",
     "shell.execute_reply.started": "2023-06-10T08:55:10.943379Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.10/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
      "  warnings.warn(\n",
      "/opt/conda/lib/python3.10/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet34_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet34_Weights.DEFAULT` to get the most up-to-date weights.\n",
      "  warnings.warn(msg)\n",
      "Downloading: \"https://download.pytorch.org/models/resnet34-b627a593.pth\" to /root/.cache/torch/hub/checkpoints/resnet34-b627a593.pth\n",
      "100%|██████████| 83.3M/83.3M [00:00<00:00, 102MB/s] \n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>error_rate</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.472346</td>\n",
       "      <td>0.341487</td>\n",
       "      <td>0.108931</td>\n",
       "      <td>01:08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>error_rate</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.443030</td>\n",
       "      <td>0.287417</td>\n",
       "      <td>0.092016</td>\n",
       "      <td>01:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.268207</td>\n",
       "      <td>0.247478</td>\n",
       "      <td>0.079838</td>\n",
       "      <td>01:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.127706</td>\n",
       "      <td>0.217290</td>\n",
       "      <td>0.066306</td>\n",
       "      <td>01:05</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn = vision_learner(dls, resnet34, metrics=error_rate)\n",
    "learn.fine_tune(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "91095764",
   "metadata": {},
   "source": [
    "We could try a better model, based on [this analysis](https://www.kaggle.com/code/jhoward/which-image-models-are-best/). The convnext models work great!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "d20f8a44",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-06-10T07:08:36.748372Z",
     "start_time": "2023-06-10T07:08:36.743317Z"
    },
    "execution": {
     "iopub.execute_input": "2023-06-10T08:59:37.731453Z",
     "iopub.status.busy": "2023-06-10T08:59:37.729116Z",
     "iopub.status.idle": "2023-06-10T08:59:37.744568Z",
     "shell.execute_reply": "2023-06-10T08:59:37.743668Z",
     "shell.execute_reply.started": "2023-06-10T08:59:37.731411Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['convnext_atto',\n",
       " 'convnext_atto_ols',\n",
       " 'convnext_base',\n",
       " 'convnext_femto',\n",
       " 'convnext_femto_ols',\n",
       " 'convnext_large',\n",
       " 'convnext_large_mlp',\n",
       " 'convnext_nano',\n",
       " 'convnext_nano_ols',\n",
       " 'convnext_pico',\n",
       " 'convnext_pico_ols',\n",
       " 'convnext_small',\n",
       " 'convnext_tiny',\n",
       " 'convnext_tiny_hnf',\n",
       " 'convnext_xlarge',\n",
       " 'convnext_xxlarge',\n",
       " 'convnextv2_atto',\n",
       " 'convnextv2_base',\n",
       " 'convnextv2_femto',\n",
       " 'convnextv2_huge',\n",
       " 'convnextv2_large',\n",
       " 'convnextv2_nano',\n",
       " 'convnextv2_pico',\n",
       " 'convnextv2_small',\n",
       " 'convnextv2_tiny']"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "timm.list_models('convnext*')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "03ba90b1",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-06-10T08:41:49.062348Z",
     "start_time": "2023-06-10T07:08:36.749543Z"
    },
    "execution": {
     "iopub.execute_input": "2023-06-10T08:59:37.747750Z",
     "iopub.status.busy": "2023-06-10T08:59:37.747399Z",
     "iopub.status.idle": "2023-06-10T09:04:07.470617Z",
     "shell.execute_reply": "2023-06-10T09:04:07.469396Z",
     "shell.execute_reply.started": "2023-06-10T08:59:37.747715Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.10/site-packages/timm/models/_factory.py:114: UserWarning: Mapping deprecated model name convnext_tiny_in22k to current convnext_tiny.fb_in22k.\n",
      "  model = create_fn(\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5ed0d349cdbf43a2b7bfd90b5d8d4a61",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading model.safetensors:   0%|          | 0.00/178M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>error_rate</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.130062</td>\n",
       "      <td>0.274265</td>\n",
       "      <td>0.083221</td>\n",
       "      <td>01:04</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>error_rate</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.284376</td>\n",
       "      <td>0.228118</td>\n",
       "      <td>0.069012</td>\n",
       "      <td>01:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.202384</td>\n",
       "      <td>0.177204</td>\n",
       "      <td>0.052774</td>\n",
       "      <td>01:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.132728</td>\n",
       "      <td>0.159799</td>\n",
       "      <td>0.050068</td>\n",
       "      <td>01:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn = vision_learner(dls, 'convnext_tiny_in22k', metrics=error_rate).to_fp16()\n",
    "learn.fine_tune(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "8f4285ca",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2023-06-10T08:41:49.084075Z",
     "start_time": "2023-06-10T08:41:49.084068Z"
    },
    "execution": {
     "iopub.execute_input": "2023-06-10T09:04:07.477615Z",
     "iopub.status.busy": "2023-06-10T09:04:07.475375Z",
     "iopub.status.idle": "2023-06-10T09:04:07.993243Z",
     "shell.execute_reply": "2023-06-10T09:04:07.992276Z",
     "shell.execute_reply.started": "2023-06-10T09:04:07.477573Z"
    }
   },
   "outputs": [],
   "source": [
    "learn.export('model.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "7dcf5a26",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-06-10T09:09:46.142957Z",
     "iopub.status.busy": "2023-06-10T09:09:46.142431Z",
     "iopub.status.idle": "2023-06-10T09:09:47.355923Z",
     "shell.execute_reply": "2023-06-10T09:09:47.354680Z",
     "shell.execute_reply.started": "2023-06-10T09:09:46.142898Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/kaggle/working\n"
     ]
    }
   ],
   "source": [
    "!pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "d02b1a04",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-06-10T09:09:47.622030Z",
     "iopub.status.busy": "2023-06-10T09:09:47.621603Z",
     "iopub.status.idle": "2023-06-10T09:09:48.832629Z",
     "shell.execute_reply": "2023-06-10T09:09:48.831425Z",
     "shell.execute_reply.started": "2023-06-10T09:09:47.621991Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__notebook_source__.ipynb  model.pkl\n"
     ]
    }
   ],
   "source": [
    "!ls"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "63f2fb53",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-06-10T09:13:00.036924Z",
     "iopub.status.busy": "2023-06-10T09:13:00.036464Z",
     "iopub.status.idle": "2023-06-10T09:13:00.045917Z",
     "shell.execute_reply": "2023-06-10T09:13:00.044440Z",
     "shell.execute_reply.started": "2023-06-10T09:13:00.036887Z"
    }
   },
   "source": [
    "<a href=\"model.pkl\"> Download model.pkl File </a>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "68065602",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "fastai_related",
   "language": "python",
   "name": "fastai_related"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.5"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": false,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "384px"
   },
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}