{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "def9fcb6",
   "metadata": {},
   "source": [
    "# Prepare Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8bb0209c-63c4-4601-894c-0ded8f4db2e6",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "d:\\sumobot\\sumobot_ml\\venv\\Lib\\site-packages\\tqdm\\auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "import sys, os\n",
    "sys.path.append(os.path.abspath(\"..\"))\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import glob\n",
    "from dataset_helper import get_dataset, get_dataset_dir\n",
    "\n",
    "\n",
    "# Amount of dataset lines that will be compiled and converted to dataset.jsonl. \n",
    "# If -1, use all lines.\n",
    "# max_dataset=100\n",
    "max_dataset=-1\n",
    "\n",
    "output_onnx_name = \"ml.onnx\"\n",
    "output_labels_name = \"ml_labels.json\"\n",
    "\n",
    "# Load & process data\n",
    "\n",
    "df, dir = get_dataset(inside_arena=True)\n",
    "\n",
    "# df.to_csv(f\"{get_dataset_dir()}/merged.csv\", index=False)\n",
    "\n",
    "if max_dataset>-1:\n",
    "    df = df.sample(max_dataset)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "80959887",
   "metadata": {},
   "source": [
    "# Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "817107a1-86f4-49de-a366-f1e80536ecef",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import tf2onnx\n",
    "\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.impute import SimpleImputer\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.layers import Input, Dense, BatchNormalization\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from tensorflow.keras.losses import CategoricalCrossentropy\n",
    "\n",
    "# Load & process data\n",
    "from dataset_helper import get_dataset\n",
    "\n",
    "df, dir = get_dataset(inside_arena=True)\n",
    "\n",
    "if max_dataset>-1:\n",
    "    df = df.sample(max_dataset)\n",
    "\n",
    "features = [\n",
    "    \"BotPosX\", \n",
    "    \"BotPosY\", \n",
    "    \"BotRot\", \n",
    "    # \"BotLinv\",\n",
    "    # \"BotAngv\", \n",
    "    # \"BotIsDashActive\",\n",
    "    # \"BotIsSkillActive\", \n",
    "    # \"BotIsOutFromArena\",\n",
    "    \"EnemyBotPosX\", \n",
    "    \"EnemyBotPosY\", \n",
    "    \"EnemyBotRot\",\n",
    "    # \"EnemyBotLinv\",\n",
    "    # \"EnemyBotAngv\", \n",
    "    # \"EnemyBotIsDashActive\",\n",
    "    # \"EnemyBotIsSkillActive\", \n",
    "    # \"EnemyBotIsOutFromArena\",\n",
    "]\n",
    "\n",
    "X = df[features]\n",
    "imputer = SimpleImputer(strategy=\"mean\")\n",
    "X = imputer.fit_transform(X)\n",
    "\n",
    "# Encode label\n",
    "le = LabelEncoder()\n",
    "y_action = le.fit_transform(df[\"Name\"])\n",
    "y_duration = df[\"Duration\"].values.astype(\"float32\")\n",
    "\n",
    "# One-hot encoding for action\n",
    "y_action_cat = to_categorical(y_action)\n",
    "\n",
    "# Split\n",
    "X_train, X_test, y_action_train, y_action_test, y_duration_train, y_duration_test, df_train, df_val = train_test_split(\n",
    "    X, y_action_cat, y_duration, df, test_size=0.2, random_state=42\n",
    ")\n",
    "\n",
    "# Build model\n",
    "inputs = Input(shape=(X.shape[1], ))\n",
    "x = Dense(256, activation='relu')(inputs)\n",
    "x = BatchNormalization()(x)\n",
    "x = Dense(128, activation='relu')(x)\n",
    "x = Dense(64, activation='relu')(x)\n",
    "x = Dense(32, activation='relu')(x)\n",
    "\n",
    "output_action = Dense(y_action_cat.shape[1], activation='softmax', name=\"action\")(x)\n",
    "output_duration = Dense(1, activation='linear', name=\"duration\")(x)\n",
    "\n",
    "loss_action = CategoricalCrossentropy(label_smoothing=0.1)\n",
    "\n",
    "# Compile model\n",
    "model = Model(inputs=inputs, outputs=[output_action, output_duration])\n",
    "model.compile(\n",
    "    optimizer=Adam(learning_rate=0.0001),\n",
    "    loss={\"action\": loss_action, \"duration\": \"mae\"},\n",
    "    metrics={'action': 'accuracy', 'duration': 'mae'},\n",
    "    weighted_metrics={'action': 'accuracy', 'duration': 'mae'}\n",
    ")\n",
    "\n",
    "# Early stopping\n",
    "early_stop = EarlyStopping(\n",
    "    monitor='val_loss',\n",
    "    patience=10,\n",
    "    min_delta=0.001,\n",
    "    mode='min',\n",
    "    restore_best_weights=True,\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "# Train\n",
    "model.fit(X_train, {\"action\": y_action_train, \"duration\": y_duration_train},\n",
    "          validation_data=(X_test, {'action': y_action_test, 'duration': y_duration_test}),\n",
    "          epochs=100,\n",
    "          batch_size=512,\n",
    "          callbacks=[early_stop],\n",
    "          )\n",
    "\n",
    "# Predict\n",
    "pred_action_prob, pred_duration = model.predict(X_test)\n",
    "pred_action = np.argmax(pred_action_prob, axis=1)\n",
    "true_action = np.argmax(y_action_test, axis=1)\n",
    "\n",
    "# Evaluation\n",
    "print(classification_report(true_action, pred_action, target_names=le.classes_))\n",
    "\n",
    "# Convert the model\n",
    "spec = (tf.TensorSpec((None, X.shape[1]), tf.float32, name=\"input\"),)\n",
    "onnx_model, _ = tf2onnx.convert.from_keras(model, input_signature=spec, opset=13)\n",
    "\n",
    "# Save to file\n",
    "with open(output_onnx_name, \"wb\") as f:\n",
    "    f.write(onnx_model.SerializeToString())\n",
    "\n",
    "print(f\"Model saved to {output_onnx_name}\")\n",
    "\n",
    "class_labels = le.classes_.tolist()\n",
    "\n",
    "# Optional: Save labels to JSON\n",
    "with open(output_labels_name, \"w\") as f:\n",
    "    json.dump(class_labels, f)\n",
    "\n",
    "print(f\"Exported label encoder classes to {output_labels_name}\")\n",
    "print(class_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "33783b51",
   "metadata": {},
   "source": [
    "# Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2814352b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import onnxruntime as ort\n",
    "import numpy as np\n",
    "import joblib\n",
    "\n",
    "# Load ONNX session\n",
    "session = ort.InferenceSession(output_onnx_name)\n",
    "\n",
    "# Load label encoder\n",
    "le = joblib.load(output_labels_name)\n",
    "\n",
    "# 1 sample input (bisa ambil dari X_test atau manual)\n",
    "sample = np.array([[\n",
    "    -3.30086637,    # BotPosX\n",
    "    2.01016259,   # BotPosY\n",
    "    # 0,   # BotLinv\n",
    "    # 0,   # BotAngv\n",
    "    270,   # BotRot\n",
    "    3.69913363,    # EnemyBotPosX\n",
    "    2.01016259,   # EnemyBotPosY\n",
    "    # 0,   # EnemyBotLinv\n",
    "    # 0,  # EnemyBotAngv\n",
    "    90   # EnemyBotRot\n",
    "]], dtype=np.float32)\n",
    "\n",
    "# Get input & output names\n",
    "input_name = session.get_inputs()[0].name\n",
    "output_names = [o.name for o in session.get_outputs()]  # action, duration\n",
    "\n",
    "# Run inference\n",
    "outputs = session.run(output_names, {input_name: sample})\n",
    "pred_action_probs = outputs[0]\n",
    "pred_duration = outputs[1]\n",
    "\n",
    "# Decode predicted action\n",
    "pred_action_index = np.argmax(pred_action_probs)\n",
    "pred_action_label = le.inverse_transform([pred_action_index])[0]\n",
    "\n",
    "print(f\"Predicted action: {pred_action_label}\")\n",
    "print(f\"Predicted duration: {pred_duration[0][0]:.4f}\")\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "venv",
   "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": "3.12.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}