{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "82baf493-aca3-40ae-8d2f-33adafecb6a9",
   "metadata": {},
   "outputs": [],
   "source": [
    "#|default_exp app"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5fec5815-2555-4b0d-bd1c-a77a7fbdeda7",
   "metadata": {},
   "source": [
    "# Digit parser\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "2c3da714-bd9c-4b8f-ae28-980f8dea239c",
   "metadata": {},
   "outputs": [],
   "source": [
    "#|export\n",
    "import torch\n",
    "import numpy as np\n",
    "import gradio as gr\n",
    "from PIL import Image\n",
    "from pathlib import Path\n",
    "import sys\n",
    "np.set_printoptions(threshold=sys.maxsize)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "5664caad-faca-489c-a8ab-74514aa7d706",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[2mAudited \u001b[1m1 package\u001b[0m \u001b[2min 2ms\u001b[0m\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "!uv pip install torchmetrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bff78822-ebd1-4f5f-a765-cb0df804a29b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "* Running on local URL:  http://127.0.0.1:7862\n",
      "\n",
      "To create a public link, set `share=True` in `launch()`.\n",
      "Keyboard interruption in main thread... closing server.\n"
     ]
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#|export\n",
    "from lenet import LeNet5\n",
    "# Allowlist the custom class\n",
    "MODEL_PATH = Path(\"models/lenet5-cpu.pt\")\n",
    "model = torch.load(MODEL_PATH, weights_only=False)\n",
    "model.eval()\n",
    "\n",
    "def predict(img):\n",
    "    # Create a new image with a white background\n",
    "    background = Image.new(\"L\", (28, 28), 255)\n",
    "\n",
    "    # Resize the input image\n",
    "    img_pil = img[\"composite\"].resize((28, 28))\n",
    "\n",
    "    # Paste the resized image onto the white background\n",
    "    background.paste(img_pil, (0, 0), img_pil)\n",
    "    \n",
    "    # Convert to numpy\n",
    "    img_array = np.array(background)\n",
    "    \n",
    "    # Invert colors (MNIST has white digits on black)\n",
    "    img_array = 255 - img_array\n",
    "\n",
    "    # Create a displayable version of the inverted image (what the model actually sees)\n",
    "    inverted_debug = img_array.astype(np.uint8)\n",
    "\n",
    "    img_tensor = torch.tensor(img_array, dtype=torch.float32)    \n",
    "    img_tensor = img_tensor.unsqueeze(0).unsqueeze(0)  # Add channel and batch dimensions\n",
    "\n",
    "    # Debug: Print the shape and values of the input tensor\n",
    "    print(f\"Input tensor shape: {img_tensor.shape}\")\n",
    "    print(f\"Input tensor values: {img_tensor}\")\n",
    "\n",
    "    with torch.no_grad():\n",
    "        output = model(img_tensor)\n",
    "        probabilities = torch.nn.functional.softmax(output, dim=1)[0]\n",
    "\n",
    "    print(f\"Output shape: {output.shape}\")\n",
    "    print(f\"Probabilities shape: {probabilities.shape}\")\n",
    "    print(f\"Probabilities: {probabilities}\")\n",
    "\n",
    "    # Create dictionary of label: probability for Gradio Label output\n",
    "    return {str(i): float(prob) for i, prob in enumerate(probabilities)}, inverted_debug\n",
    "\n",
    "image = gr.Sketchpad(type=\"pil\", sources=(), canvas_size=(280,280), brush=gr.Brush(colors=[\"#000000\"], color_mode=\"fixed\", default_size=20), layers=False, transforms=[])\n",
    "label = gr.Label()\n",
    "processed_image = gr.Image(label=\"What the Model Sees (28x28)\")\n",
    "intf = gr.Interface(title=\"Draw a digit\", description=\"And let me identify it for you...\", fn=predict, inputs=image, outputs=[label, processed_image], clear_btn=None)\n",
    "intf.launch(inline=False, debug=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf53a6ec-86bf-44cb-baaa-011f21f5869e",
   "metadata": {},
   "source": [
    "## Export"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "c35ecd80-c0a1-421a-9dd2-04cca2d4c461",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[2mUsing Python 3.12.7 environment at: /home/afonso/git/private/pytorch-tutorial/.venv\u001b[0m\n",
      "\u001b[2mAudited \u001b[1m1 package\u001b[0m \u001b[2min 34ms\u001b[0m\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "!uv pip install nbdev\n",
    "from nbdev.export import nb_export"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "de31d563-3696-45ba-9100-06c93072508c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Exported\n"
     ]
    }
   ],
   "source": [
    "nb_export('app.ipynb', './')\n",
    "print(\"Exported\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1a443132-c4ec-4990-89c3-9a6320d14640",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3058daee-595f-4829-ae93-a38bebdc4030",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pytorch-tutorial",
   "language": "python",
   "name": "pytorch-tutorial"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}