Inference.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "W0130 21:56:44.426367 140034566559552 word.py:92] Falling back to default tokenizer, the NLTK's `TreebankWordTokenizer()`.\n",
      "I0130 21:56:44.583137 140034566559552 dictionary.py:209] adding document #0 to Dictionary(0 unique tokens: [])\n",
      "I0130 21:56:44.584788 140034566559552 dictionary.py:216] built Dictionary(12 unique tokens: ['computer', 'human', 'interface', 'response', 'survey']...) from 9 documents (total 29 corpus positions)\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "import parmap\n",
    "import pandas as pd\n",
    "from tqdm import tqdm,tqdm_notebook\n",
    "from models.tokenization import *\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "from models.model_architecture import *\n",
    "from models.train_eval import *\n",
    "from models.model_utils import *\n",
    "from models.tokenization import *\n",
    "parent_path='Data/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "params={'model_path':'xlm-roberta-base',\n",
    "        'preprocess_doc':False,\n",
    "        'max_length':256,\n",
    "        'batch_size':16,\n",
    "        'hidden_size':128,\n",
    "        'weights':[1.0,1.0],\n",
    "        'load_saved':True,\n",
    "        'seq_model':'lstm',\n",
    "        'data_path':parent_path+'fear_speech_data.json',\n",
    "        'max_sentences_per_doc':5,\n",
    "        'transformer_type':'normal_transformer',\n",
    "        'take_tokens_from':'last',\n",
    "        'device':'cpu',\n",
    "        'learning_rate':2e-5,\n",
    "        'epsilon':1e-8,\n",
    "        'random_seed':2020,\n",
    "        'epochs':20,\n",
    "        'max_memory':0.6,\n",
    "        'freeze_bert':False\n",
    "       }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def softmax(x):\n",
    "    \"\"\"Compute softmax values for each sets of scores in x.\"\"\"\n",
    "    e_x = np.exp(x - np.max(x))\n",
    "    temp=e_x / e_x.sum(axis=0) # only difference\n",
    "    \n",
    "    if np.isnan(temp).any()==True:\n",
    "        return [0.0,1.0,0.0]\n",
    "    else:\n",
    "        return temp\n",
    "\n",
    "def return_dataframe(path):\n",
    "    with open(path, 'r') as fp:\n",
    "        json_data=json.load(fp)\n",
    "    text_tuples=[]\n",
    "    for key in json_data.keys():\n",
    "        element = json_data[key]\n",
    "        count_fearspeech=element['annotation_list'].count('Fear speech')\n",
    "        count_normal=element['annotation_list'].count('Normal')\n",
    "\n",
    "        if(count_fearspeech>count_normal):\n",
    "            one_fear_speech=1\n",
    "        else:\n",
    "            one_fear_speech=0\n",
    "        text_tuples.append([element['message_text'],one_fear_speech])\n",
    "    annotated_df = pd.DataFrame(text_tuples,columns=['message_text','one_fear_speech'])\n",
    "    return annotated_df    \n",
    "    \n",
    "\n",
    "class modelPred():\n",
    "    def __init__(self,params,fold):\n",
    "        self.params=params\n",
    "        if torch.cuda.is_available() and self.params['device']=='cuda':    \n",
    "            # Tell PyTorch to use the GPU.    \n",
    "            self.device = torch.device(\"cuda\")\n",
    "            deviceID = get_gpu(self.params)\n",
    "            torch.cuda.set_device(deviceID[0])\n",
    "        else:\n",
    "            print('Since you dont want to use GPU, using the CPU instead.')\n",
    "            self.device = torch.device(\"cpu\")\n",
    "        \n",
    "        \n",
    "        self.model=select_transformer_model(params['transformer_type'],params['model_path'],params,count_skf)\n",
    "       \n",
    "        if(self.params[\"device\"]=='cuda'):\n",
    "            self.model.cuda()\n",
    "        self.model.eval()\n",
    "    \n",
    "    def return_probab(self,sentences_list):\n",
    "        \"\"\"Input: should be a list of sentences\"\"\"\n",
    "        \"\"\"Ouput: probablity values\"\"\"\n",
    "        params=self.params\n",
    "        device = self.device\n",
    "\n",
    "        X_test,X_test_mask= encode_sent(sentences_list,params,tokenizer)\n",
    "        y_test=np.zeros((len(X_test),))\n",
    "        test_dataloader=return_dataloader_sent(X_test,X_test_mask,y_test,params,is_train=False)\n",
    "        # Put the model in evaluation mode--the dropout layers behave differently\n",
    "        # during evaluation.\n",
    "        # Tracking variables\n",
    "        \n",
    "\n",
    "        print(\"Running eval on test data...\")\n",
    "        t0 = time.time()\n",
    "        true_labels=[]\n",
    "        pred_labels=[]\n",
    "        logits_all=[]\n",
    "        input_mask_all=[]\n",
    "\n",
    "        # Evaluate data for one epoch\n",
    "        for step, batch in tqdm(enumerate(test_dataloader),total=len(test_dataloader)):\n",
    "\n",
    "            # Progress update every 40 batches.\n",
    "            if step % 40 == 0 and not step == 0:\n",
    "                # Calculate elapsed time in minutes.\n",
    "                elapsed = format_time(time.time() - t0)\n",
    "\n",
    "\n",
    "            # `batch` contains three pytorch tensors:\n",
    "            #   [0]: input ids \n",
    "            #   [1]: attention vals\n",
    "            #   [2]: attention mask\n",
    "            #   [3]: labels \n",
    "            b_input_ids = batch[0].to(device)\n",
    "            b_att_val = batch[1].to(device)\n",
    "            b_input_mask = batch[2].to(device)\n",
    "            b_labels = batch[3].to(device)\n",
    "\n",
    "\n",
    "            # (source: https://stackoverflow.com/questions/48001598/why-do-we-need-to-call-zero-grad-in-pytorch)\n",
    "            #model.zero_grad()        \n",
    "            outputs = self.model(b_input_ids,\n",
    "                attention_mask=b_input_mask, \n",
    "                labels=None)\n",
    "            logits = outputs\n",
    "\n",
    "            # Move logits and labels to CPU\n",
    "            logits = logits.detach().cpu().numpy()\n",
    "            label_ids = b_labels.detach().cpu().numpy()\n",
    "\n",
    "            pred_labels+=list(np.argmax(logits, axis=1).flatten())\n",
    "            true_labels+=list(label_ids.flatten())\n",
    "            logits_all+=list(logits)\n",
    "            input_mask_all+=list(batch[2].detach().cpu().numpy())\n",
    "\n",
    "\n",
    "        logits_all_final=[]\n",
    "        for logits in logits_all:\n",
    "            logits_all_final.append(list(softmax(logits)))\n",
    "\n",
    "\n",
    "        return np.array(logits_all_final)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/punyajoy/.conda/envs/punyajoy_gpu/lib/python3.7/site-packages/ipykernel_launcher.py:8: TqdmDeprecationWarning: This function will be removed in tqdm==5.0.0\n",
      "Please use `tqdm.notebook.tqdm` instead of `tqdm.tqdm_notebook`\n",
      "  \n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b843123503a942928444097473faa078",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=4782.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Loading tokenizer...\n",
      "Since you dont want to use GPU, using the CPU instead.\n",
      "======== Fold 0 / 5 ========\n",
      "[870 872 873 874 875] [0 1 2 3 4]\n",
      "\n",
      "Since you dont want to use GPU, using the CPU instead.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/punyajoy/.conda/envs/punyajoy_gpu/lib/python3.7/site-packages/sklearn/model_selection/_split.py:296: FutureWarning: Setting a random_state has no effect since shuffle is False. This will raise an error in 0.24. You should leave random_state to its default (None), or set shuffle=True.\n",
      "  FutureWarning\n",
      "  0%|          | 0/60 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 60/60 [12:54<00:00, 12.91s/it]\n",
      "100%|██████████| 957/957 [00:00<00:00, 100946.83it/s]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:42<00:00, 10.64s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:49<00:00, 12.26s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:50<00:00, 12.69s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:44<00:00, 11.12s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:48<00:00, 12.16s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:48<00:00, 12.06s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:41<00:00, 10.47s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:49<00:00, 12.33s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:46<00:00, 11.53s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:49<00:00, 12.32s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:42<00:00, 10.56s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:46<00:00, 11.74s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:48<00:00, 12.14s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:56<00:00, 14.08s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:22<00:00, 20.69s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:18<00:00, 19.60s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:46<00:00, 11.61s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:03<00:00, 15.79s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:59<00:00, 14.82s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:04<00:00, 16.13s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:05<00:00, 16.36s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:57<00:00, 14.42s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:59<00:00, 14.94s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:03<00:00, 15.87s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:03<00:00, 15.90s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:07<00:00, 16.86s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:12<00:00, 18.07s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:38<00:00,  9.60s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [16:16<00:00, 244.01s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:46<00:00, 11.70s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:32<00:00,  8.20s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:33<00:00,  8.39s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:22<00:00,  5.60s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:18<00:00,  4.52s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:22<00:00,  5.53s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:44<00:00, 11.13s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:05<00:00, 16.45s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:58<00:00, 14.63s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [01:00<00:00, 15.02s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [02:15<00:00, 33.86s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "======== Fold 1 / 5 ========\n",
      "[0 1 2 3 4] [870 872 873 874 875]\n",
      "\n",
      "Since you dont want to use GPU, using the CPU instead.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      "  0%|          | 0/60 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 60/60 [02:44<00:00,  2.74s/it]\n",
      "100%|██████████| 957/957 [00:00<00:00, 102462.00it/s]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:10<00:00,  2.70s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.07s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:09<00:00,  2.39s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.13s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:16<00:00,  4.01s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:09<00:00,  2.36s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.17s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.20s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.13s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  2.00s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.08s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.02s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.02s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.19s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.97s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.06s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.07s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.13s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.93s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.04s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.07s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.99s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.00s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.96s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:09<00:00,  2.32s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.01s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.90s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.98s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.05s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.12s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.10s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.20s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:09<00:00,  2.33s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.20s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.91s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.89s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.94s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  2.00s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.92s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.88s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "======== Fold 2 / 5 ========\n",
      "[0 1 2 3 4] [1771 1772 1773 1774 1775]\n",
      "\n",
      "Since you dont want to use GPU, using the CPU instead.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      "  0%|          | 0/60 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 60/60 [02:03<00:00,  2.06s/it]\n",
      "100%|██████████| 956/956 [00:00<00:00, 26252.16it/s]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.93s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.86s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.04s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.91s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.81s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.80s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.89s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.87s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:06<00:00,  1.75s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.80s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:09<00:00,  2.33s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:10<00:00,  2.61s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:09<00:00,  2.29s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.13s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.09s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.20s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:09<00:00,  2.32s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:09<00:00,  2.28s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.10s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.20s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.25s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.92s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.82s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.77s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.78s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.83s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.93s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.80s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.79s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.79s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.86s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.82s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.83s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.79s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.81s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.85s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.80s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.76s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.84s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.81s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "======== Fold 3 / 5 ========\n",
      "[0 1 2 3 4] [2757 2758 2759 2761 2762]\n",
      "\n",
      "Since you dont want to use GPU, using the CPU instead.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      "  0%|          | 0/60 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 60/60 [02:06<00:00,  2.10s/it]\n",
      "100%|██████████| 956/956 [00:00<00:00, 223061.56it/s]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.82s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.90s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.79s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:06<00:00,  1.74s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.83s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.90s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.82s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.85s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.77s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.77s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.79s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.80s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.79s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:07<00:00,  1.83s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:08<00:00,  2.01s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:27<00:00,  6.97s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:39<00:00,  9.76s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:42<00:00, 10.54s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:35<00:00,  9.00s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:34<00:00,  8.55s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:36<00:00,  9.09s/it]\n",
      "  0%|          | 0/4 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running eval on test data...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|█████     | 2/4 [00:20<00:19,  9.58s/it]"
     ]
    }
   ],
   "source": [
    "import random \n",
    "from lime.lime_text import LimeTextExplainer\n",
    "from transformers import AutoTokenizer, AutoModelForSequenceClassification\n",
    "annotated_df=return_dataframe(params['data_path'])\n",
    "\n",
    "#### Preprocessing\n",
    "params_preprocess={'remove_numbers': True, 'remove_emoji': True, 'remove_stop_words': False, 'tokenize': False}\n",
    "list_sents=[preprocess_sent(ele,params=params_preprocess) for ele in tqdm_notebook(annotated_df['message_text'],total=len(annotated_df))]\n",
    "\n",
    "\n",
    "#### Data_arrays\n",
    "X_0 = np.array(list_sents,dtype='object')\n",
    "y_0 = np.array(annotated_df['one_fear_speech'])\n",
    "\n",
    "\n",
    "\n",
    "if(params['weights']== [1.0,1.0]):\n",
    "    params['weights']=list(class_weight.compute_class_weight(\"balanced\", np.unique(y_0),y_0).astype(float))\n",
    "\n",
    "print('Loading tokenizer...')\n",
    "\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(params['model_path'])\n",
    "\n",
    "if torch.cuda.is_available() and params['device']=='cuda':    \n",
    "    # Tell PyTorch to use the GPU.    \n",
    "    device = torch.device(\"cuda\")\n",
    "    deviceID = get_gpu(params)\n",
    "    torch.cuda.set_device(deviceID[0])\n",
    "else:\n",
    "    print('Since you dont want to use GPU, using the CPU instead.')\n",
    "    device = torch.device(\"cpu\")\n",
    "\n",
    "\n",
    "list_dict=[]\n",
    "skf = StratifiedKFold(n_splits=5, random_state= 2020)\n",
    "count_skf=0\n",
    "\n",
    "\n",
    "explainer = LimeTextExplainer(class_names=['non fear speech','fear speech'],\n",
    "                              split_expression='\\s+',random_state=333,bow=False)\n",
    "\n",
    "y_ground_all=[]\n",
    "y_pred_all=[]\n",
    "\n",
    "for train_index, test_index in skf.split(X_0, y_0):\n",
    "    print('======== Fold {:} / {:} ========'.format(count_skf,5))\n",
    "    print(train_index[0:5], test_index[0:5])\n",
    "    count_skf+=1\n",
    "    print(\"\")\n",
    "    X_train, X_test = X_0[train_index], X_0[test_index]\n",
    "    y_train, y_test = y_0[train_index], y_0[test_index]\n",
    "    \n",
    "    \n",
    "    model_cl=modelPred(params,count_skf)\n",
    "    \n",
    "    \n",
    "    logits_all_final=model_cl.return_probab(X_test)\n",
    "    \n",
    "    index_ground_nfs=[]\n",
    "    index_ground_fs=[]\n",
    "    \n",
    "    \n",
    "    for i in tqdm(range(len(y_test))):\n",
    "        pred_id=np.argmax(logits_all_final[i])\n",
    "        ground_label=y_test[i]\n",
    "        y_ground_all.append(ground_label)\n",
    "        y_pred_all.append(pred_id)\n",
    "        if(pred_id!=ground_label):\n",
    "            if(ground_label==1):\n",
    "                index_ground_fs.append(i)\n",
    "            else:\n",
    "                index_ground_nfs.append(i)\n",
    "    \n",
    "    for i in random.sample(index_ground_fs,20):\n",
    "        temp={}\n",
    "        exp = explainer.explain_instance(X_test[i], model_cl.return_probab, num_features=6, top_labels=2,num_samples=50)\n",
    "        pred_id=np.argmax(exp.predict_proba)\n",
    "        ground_label=y_test[i]\n",
    "        temp[\"text\"]=X_test[i]\n",
    "        temp[\"annotation\"]=ground_label\n",
    "        temp[\"classification\"]=pred_id\n",
    "        temp[\"classification_scores\"]={\"non fear speech\":exp.predict_proba[0],\"fear speech\":exp.predict_proba[1]}\n",
    "        temp[\"text_explained\"]=exp.as_list()\n",
    "        list_dict.append(temp)\n",
    "    \n",
    "    for i in random.sample(index_ground_nfs,20):\n",
    "        temp={}\n",
    "        exp = explainer.explain_instance(X_test[i], model_cl.return_probab, num_features=6, top_labels=2,num_samples=50)\n",
    "        pred_id=np.argmax(exp.predict_proba)\n",
    "        ground_label=y_test[i]\n",
    "        temp[\"text\"]=X_test[i]\n",
    "        temp[\"annotation\"]=ground_label\n",
    "        temp[\"classification\"]=pred_id\n",
    "        temp[\"classification_scores\"]={\"non fear speech\":exp.predict_proba[0],\"fear speech\":exp.predict_proba[1]}\n",
    "        temp[\"text_explained\"]=exp.as_list()\n",
    "        list_dict.append(temp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix,make_scorer\n",
    "from sklearn.metrics import f1_score, accuracy_score, recall_score, precision_score\n",
    "from sklearn.metrics import classification_report, precision_recall_fscore_support"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "testf1=f1_score(y_ground_all, y_pred_all, average='macro')\n",
    "testacc=accuracy_score(y_ground_all,y_pred_all)\n",
    "testrocauc=roc_auc_score(y_ground_all, y_pred_all,average='macro')\n",
    "testprecision=precision_score(y_ground_all, y_pred_all, average='binary')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val_fscore 0.6531715907724819\n",
      "val_accuracy 0.7258469259723965\n",
      "val_rocauc 0.6693842497257558\n",
      "val_precision 0.44176430048242593\n"
     ]
    }
   ],
   "source": [
    "print('val_fscore',testf1)\n",
    "print('val_accuracy',testacc)\n",
    "print('val_rocauc',testrocauc)\n",
    "print('val_precision',testprecision)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "\n",
    "class NumpyEncoder(json.JSONEncoder):\n",
    "    \"\"\" Special json encoder for numpy types \"\"\"\n",
    "    def default(self, obj):\n",
    "        if isinstance(obj, (np.int_, np.intc, np.intp, np.int8,\n",
    "            np.int16, np.int32, np.int64, np.uint8,\n",
    "            np.uint16, np.uint32, np.uint64)):\n",
    "            return int(obj)\n",
    "        elif isinstance(obj, (np.float_, np.float16, np.float32, \n",
    "            np.float64)):\n",
    "            return float(obj)\n",
    "        elif isinstance(obj,(np.ndarray,)): #### This is the fix\n",
    "            return obj.tolist()\n",
    "        return json.JSONEncoder.default(self, obj)\n",
    "\n",
    "\n",
    "with open('lime_explanation.txt', 'w') as fp:\n",
    "        \n",
    "    for ele in list_dict:\n",
    "        fp.write(\"====================Start=======================\\n\")\n",
    "        fp.write(json.dumps(ele,cls=NumpyEncoder,ensure_ascii=False))\n",
    "        fp.write(\"====================End=======================\\n\")\n",
    "        \n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(list_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[2979  661]\n",
      " [ 505  637]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix#Fit the model\n",
    "cf_matrix = confusion_matrix(y_ground_all,y_pred_all)\n",
    "print(cf_matrix)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 504x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "matplotlib.rcParams['figure.figsize'] = (7, 5)\n",
    "sns.set(context='paper',style='white',font_scale=2.0)\n",
    "\n",
    "#labels = [‘True Neg’,’False Pos’,’False Neg’,’True Pos’]\n",
    "categories = ['NFS', 'FS']\n",
    "make_confusion_matrix(cf_matrix,\n",
    "                      categories=categories, \n",
    "                      cmap='RdYlBu_r',\n",
    "                      sum_stats=False)\n",
    "plt.savefig('../Fear_speech_results/confusion_matrix.pdf',bbox_inches='tight',dpi=400)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:.conda-punyajoy_gpu] *",
   "language": "python",
   "name": "conda-env-.conda-punyajoy_gpu-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}