{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "544a588c-68ff-440f-be5c-389f1f02a0b7",
   "metadata": {},
   "source": [
    "# example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "7ef8c97c-cefd-4905-8d63-af303c412d1a",
   "metadata": {},
   "outputs": [],
   "source": [
    "MODEL_NAME = \"gaepago-20-lite\"\n",
    "DATASET_NAME = \"Gae8J/modeling_v1\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "044499ce-7821-4b59-9f4b-5971b6a24cce",
   "metadata": {},
   "source": [
    "## load dataset (test data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "e827e3bb-820d-46b3-b2e8-fdb97787bde1",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:datasets.builder:Found cached dataset parquet (/home/jovyan/.cache/huggingface/datasets/Gae8J___parquet/Gae8J--modeling_v1-b480c78c61a26816/0.0.0/2a3b91fbd88a2c90d1dbbb32b460cf621d31bd5b05b934492fdef7d8d6f236ec)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4438f0b33464423b92fecc698c1935e5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/3 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from datasets import load_dataset, Audio\n",
    "from transformers import AutoFeatureExtractor\n",
    "dataset = load_dataset(DATASET_NAME)\n",
    "dataset = dataset.cast_column(\"audio\", Audio(sampling_rate=16000))\n",
    "test_data = dataset['test']\n",
    "sampling_rate = test_data.features[\"audio\"].sampling_rate\n",
    "feature_extractor = AutoFeatureExtractor.from_pretrained(MODEL_NAME)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "779c547a-7e27-4481-8a66-fd9900e41964",
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoConfig\n",
    "config = AutoConfig.from_pretrained(MODEL_NAME)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "03659af7-3d90-4431-a4ea-a8d99e93602f",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "0f58cfcf-ba2d-45e4-b4e9-87df88e9dbad",
   "metadata": {},
   "outputs": [],
   "source": [
    "loaded_quantized_model = torch.jit.load(\"gaepago-20-lite/model_quant_int8.pt\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "52212656-a3e9-4bd2-ac2d-427acb5795c6",
   "metadata": {},
   "source": [
    "## 모델결과"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "3d4f5365-d6f1-4163-9c47-ce8c89e13884",
   "metadata": {},
   "outputs": [],
   "source": [
    "preds = []\n",
    "gts = []\n",
    "# quant_logits_list = []\n",
    "for i in range(len(test_data)):\n",
    "    inputs = feature_extractor(test_data[i][\"audio\"][\"array\"], sampling_rate=sampling_rate, return_tensors=\"pt\")\n",
    "    with torch.no_grad():\n",
    "        logits = loaded_quantized_model(**inputs)['logits']\n",
    "#     quant_logits_list.append(logits)\n",
    "    predicted_class_ids = torch.argmax(logits).item()\n",
    "    predicted_label = config.id2label[predicted_class_ids]\n",
    "    preds.append(predicted_label)\n",
    "    gts.append(config.id2label[test_data[i]['label']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "93b3c424-bab6-4774-915e-9e9f534f762d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "        bark     0.5556    0.6250    0.5882         8\n",
      "    growling     1.0000    0.8333    0.9091         6\n",
      "        howl     0.7500    0.8571    0.8000         7\n",
      "     panting     1.0000    0.8000    0.8889        10\n",
      "     whimper     0.3750    0.4286    0.4000         7\n",
      "\n",
      "    accuracy                         0.7105        38\n",
      "   macro avg     0.7361    0.7088    0.7172        38\n",
      "weighted avg     0.7452    0.7105    0.7224        38\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "test_performance = classification_report(gts, preds,digits=4)\n",
    "print(test_performance)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "99a3ea38-54c8-4aed-9bbf-12f98bf09dc5",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}