deprem-ml
/

adres_ner_v2_bert_128k

+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "c88f989c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "os.environ['CUDA_VISIBLE_DEVICES']='7'"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "bfdbe247",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "2023-02-26 02:35:07.275938: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
+      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
+      "2023-02-26 02:35:07.472394: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:07.472434: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.\n",
+      "2023-02-26 02:35:07.503598: E tensorflow/stream_executor/cuda/cuda_blas.cc:2981] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
+      "2023-02-26 02:35:08.603575: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:08.603678: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer_plugin.so.7'; dlerror: libnvinfer_plugin.so.7: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:08.603689: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.\n",
+      "2023-02-26 02:35:15.326595: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:15.326728: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcublas.so.11'; dlerror: libcublas.so.11: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:15.326831: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcublasLt.so.11'; dlerror: libcublasLt.so.11: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:15.327013: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcusolver.so.11'; dlerror: libcusolver.so.11: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:15.327108: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcusparse.so.11'; dlerror: libcusparse.so.11: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:15.327205: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudnn.so.8'; dlerror: libcudnn.so.8: cannot open shared object file: No such file or directory\n",
+      "2023-02-26 02:35:15.327224: W tensorflow/core/common_runtime/gpu/gpu_device.cc:1934] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.\n",
+      "Skipping registering GPU devices...\n"
+     ]
+    }
+   ],
+   "source": [
+    "from transformers import AutoTokenizer\n",
+    "import re\n",
+    "import numpy as np\n",
+    "from random import Random\n",
+    "import torch\n",
+    "import pandas as pd\n",
+    "import spacy\n",
+    "import random\n",
+    "from datasets import load_dataset\n",
+    "from transformers import (\n",
+    "    AutoModelForTokenClassification,\n",
+    "    AutoTokenizer,\n",
+    "    DataCollatorForTokenClassification,\n",
+    "    TrainingArguments,\n",
+    "    Trainer,\n",
+    "    set_seed)\n",
+    "import numpy as np\n",
+    "import datasets\n",
+    "from collections import defaultdict\n",
+    "from datasets import load_metric"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "7a916e9f",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# !pip install seqeval"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "4b0590b7",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "per_device_train_batch_size = 16\n",
+    "per_device_eval_batch_size = 32\n",
+    "num_train_epochs = 5\n",
+    "weight_decay = 0.1\n",
+    "warmup_ratio = 0.1\n",
+    "learning_rate = 5e-5\n",
+    "load_best_model_at_end = True\n",
+    "output_dir = \"../akoksal/earthquake_ner_models/\"\n",
+    "old_data_path = \"annotated_address_dataset_07022023_766train_192test/\"\n",
+    "data_path = \"deprem-private/ner_v12\"\n",
+    "cache_dir = \"../akoksal/hf_cache\"\n",
+    "saved_models_path = \"../akoksal/earthquake_ner_models/\"\n",
+    "device = \"cuda\"\n",
+    "seed = 42\n",
+    "model_names = [\"dbmdz/bert-base-turkish-cased\",\n",
+    "              \"dbmdz/electra-base-turkish-mc4-cased-discriminator\",\n",
+    "              \"dbmdz/bert-base-turkish-128k-cased\",\n",
+    "              \"dbmdz/convbert-base-turkish-cased\",\n",
+    "              \"bert-base-multilingual-cased\",\n",
+    "              \"xlm-roberta-base\"]\n",
+    "model_name = model_names[2]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "9aeb3dbe",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'dbmdz/bert-base-turkish-128k-cased'"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "model_name"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "ffeb73e4",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "set_seed(seed)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "a876c516",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "id2label = {\n",
+    "    0: \"O\",\n",
+    "    1: \"B-bina\",\n",
+    "    2: \"I-bina\",\n",
+    "    3: \"B-bulvar\",\n",
+    "    4: \"I-bulvar\",\n",
+    "    5: \"B-cadde\",\n",
+    "    6: \"I-cadde\",\n",
+    "    7: \"B-diskapino\",\n",
+    "    8: \"I-diskapino\",\n",
+    "    9: \"B-ilce\",\n",
+    "    10: \"I-ilce\",\n",
+    "    11: \"B-isim\",\n",
+    "    12: \"I-isim\",\n",
+    "    13: \"B-mahalle\",\n",
+    "    14: \"I-mahalle\",\n",
+    "    15: \"B-sehir\",\n",
+    "    16: \"I-sehir\",\n",
+    "    17: \"B-site\",\n",
+    "    18: \"I-site\",\n",
+    "    19: \"B-sokak\",\n",
+    "    20: \"I-sokak\",\n",
+    "    21: \"B-soyisim\",\n",
+    "    22: \"I-soyisim\",\n",
+    "    23: \"B-telefonno\",\n",
+    "    24: \"I-telefonno\",\n",
+    "}\n",
+    "\n",
+    "label2id = {label: idx for idx, label in id2label.items()}\n",
+    "label_names = list(label2id.keys())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "2e0caffc",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# from huggingface_hub import login\n",
+    "# login()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "c74850f9",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Some weights of the model checkpoint at dbmdz/bert-base-turkish-128k-cased were not used when initializing BertForTokenClassification: ['cls.predictions.transform.dense.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.bias', 'cls.predictions.decoder.weight', 'cls.seq_relationship.weight', 'cls.seq_relationship.bias']\n",
+      "- This IS expected if you are initializing BertForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
+      "- This IS NOT expected if you are initializing BertForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
+      "Some weights of BertForTokenClassification were not initialized from the model checkpoint at dbmdz/bert-base-turkish-128k-cased and are newly initialized: ['classifier.bias', 'classifier.weight']\n",
+      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
+     ]
+    }
+   ],
+   "source": [
+    "tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
+    "model = AutoModelForTokenClassification.from_pretrained(model_name,\n",
+    "                                                        num_labels=len(label_names),\n",
+    "                                                        id2label=id2label,\n",
+    "                                                        cache_dir=cache_dir).to(device)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "4c1fe653",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Using custom data configuration deprem-private--ner_v12-e2f61c5a18a7a738\n",
+      "Found cached dataset text (/mounts/Users/cisintern/akoksal/.cache/huggingface/datasets/deprem-private___text/deprem-private--ner_v12-e2f61c5a18a7a738/0.0.0/cb1e9bd71a82ad27976be3b12b407850fe2837d80c22c5e03a28949843a8ace2)\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "22bc5f5f97204b41b2bc5dc3b71036e1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "  0%|          | 0/3 [00:00<?, ?it/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "raw_dataset = datasets.load_dataset(\"deprem-private/ner_v12\", use_auth_token=True)\n",
+    "\n",
+    "new_dataset_json = {}\n",
+    "for split in [\"train\", \"validation\", \"test\"]:\n",
+    "    ids = []\n",
+    "    sentences = []\n",
+    "    labels = []\n",
+    "    ids = []\n",
+    "    cur_idx = 0\n",
+    "    unique_labels = set()\n",
+    "    temp_sent = []\n",
+    "    temp_labels = []\n",
+    "    for word in raw_dataset[split][\"text\"]:\n",
+    "        \n",
+    "        if word!=\"\":\n",
+    "            temp_sent.append((word.split()[0]))\n",
+    "            temp_labels.append(label2id[(word.split()[1])])\n",
+    "        else:\n",
+    "            sentences.append(temp_sent)\n",
+    "            labels.append(temp_labels)\n",
+    "            ids.append(cur_idx)\n",
+    "            cur_idx+=1\n",
+    "            temp_sent = []\n",
+    "            temp_labels = []\n",
+    "    new_dataset_json[split] = {\"tokens\":sentences, \"ner_tags\":labels, \"ids\":ids}\n",
+    "\n",
+    "dataset = datasets.DatasetDict()\n",
+    "# using your `Dict` object\n",
+    "for k,v in new_dataset_json.items():\n",
+    "    dataset[k] = datasets.Dataset.from_dict(v)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "65a66af9",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a403f5fadb3041f4b18acc7ec41a2d36",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "  0%|          | 0/1 [00:00<?, ?ba/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "e2410f6106514cfd8207d8b42748c66d",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "  0%|          | 0/1 [00:00<?, ?ba/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "227e163e07b2414da9abdbe11cb0c6bf",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "  0%|          | 0/1 [00:00<?, ?ba/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "# dataset = datasets.load_from_disk(old_data_path)\n",
+    "def tokenize_and_align_labels(examples):\n",
+    "    tokenized_inputs = tokenizer(examples[\"tokens\"], truncation=True, is_split_into_words=True)\n",
+    "\n",
+    "    labels = []\n",
+    "    for i, label in enumerate(examples[f\"ner_tags\"]):\n",
+    "        word_ids = tokenized_inputs.word_ids(batch_index=i)  # Map tokens to their respective word.\n",
+    "        previous_word_idx = None\n",
+    "        label_ids = []\n",
+    "        for word_idx in word_ids:  # Set the special tokens to -100.\n",
+    "            if word_idx is None:\n",
+    "                label_ids.append(-100)\n",
+    "            elif word_idx != previous_word_idx:  # Only label the first token of a given word.\n",
+    "                label_ids.append(label[word_idx])\n",
+    "            else:\n",
+    "                label_ids.append(-100)\n",
+    "            previous_word_idx = word_idx\n",
+    "        labels.append(label_ids)\n",
+    "\n",
+    "    tokenized_inputs[\"labels\"] = labels\n",
+    "    return tokenized_inputs\n",
+    "\n",
+    "tokenized_dataset = dataset.map(tokenize_and_align_labels, batched=True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "6b43934d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "data_collator = DataCollatorForTokenClassification(tokenizer)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "c24f52db",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/tmp/ipykernel_2652487/885599324.py:1: FutureWarning: load_metric is deprecated and will be removed in the next major version of datasets. Use 'evaluate.load' instead, from the new library 🤗 Evaluate: https://huggingface.co/docs/evaluate\n",
+      "  metric = load_metric(\"seqeval\")\n"
+     ]
+    }
+   ],
+   "source": [
+    "metric = load_metric(\"seqeval\")\n",
+    "def compute_metrics(p):\n",
+    "    predictions, labels = p\n",
+    "    predictions = np.argmax(predictions, axis=2)\n",
+    "\n",
+    "    # Remove ignored index (special tokens)\n",
+    "    true_predictions = [\n",
+    "        [label_names[p] for (p, l) in zip(prediction, label) if l != -100]\n",
+    "        for prediction, label in zip(predictions, labels)\n",
+    "    ]\n",
+    "    true_labels = [\n",
+    "        [label_names[l] for (p, l) in zip(prediction, label) if l != -100]\n",
+    "        for prediction, label in zip(predictions, labels)\n",
+    "    ]\n",
+    "\n",
+    "    results = metric.compute(predictions=true_predictions, references=true_labels)\n",
+    "    flattened_results = {\n",
+    "        \"overall_precision\": results[\"overall_precision\"],\n",
+    "        \"overall_recall\": results[\"overall_recall\"],\n",
+    "        \"overall_f1\": results[\"overall_f1\"],\n",
+    "        \"overall_accuracy\": results[\"overall_accuracy\"],\n",
+    "    }\n",
+    "    for k in results.keys():\n",
+    "        if(k not in flattened_results.keys()):\n",
+    "            flattened_results[k+\"_f1\"]=results[k][\"f1\"]\n",
+    "            flattened_results[k+\"_recall\"]=results[k][\"recall\"]\n",
+    "            flattened_results[k+\"_precision\"]=results[k][\"precision\"]\n",
+    "            flattened_results[k+\"_support\"]=results[k][\"number\"]\n",
+    "\n",
+    "    return flattened_results"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "a955fd51",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "training_args = TrainingArguments(\n",
+    "    output_dir=saved_models_path,\n",
+    "    evaluation_strategy=\"epoch\",\n",
+    "    learning_rate=learning_rate,\n",
+    "    per_device_train_batch_size=per_device_train_batch_size,\n",
+    "    per_device_eval_batch_size=per_device_eval_batch_size,\n",
+    "    num_train_epochs=num_train_epochs,\n",
+    "    warmup_ratio=warmup_ratio,\n",
+    "    weight_decay=weight_decay,\n",
+    "    run_name = \"turkish_ner\",\n",
+    "    save_strategy='epoch',\n",
+    "    logging_strategy=\"epoch\",\n",
+    "    save_total_limit=3,\n",
+    "    load_best_model_at_end=load_best_model_at_end,\n",
+    "    \n",
+    ")\n",
+    "trainer = Trainer(\n",
+    "    model=model,\n",
+    "    args=training_args,\n",
+    "    train_dataset=tokenized_dataset[\"train\"],\n",
+    "    eval_dataset=tokenized_dataset[\"validation\"],\n",
+    "    data_collator=data_collator,\n",
+    "    tokenizer=tokenizer,\n",
+    "    compute_metrics=compute_metrics\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "9f78efdc",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "The following columns in the training set don't have a corresponding argument in `BertForTokenClassification.forward` and have been ignored: tokens, ids, ner_tags. If tokens, ids, ner_tags are not expected by `BertForTokenClassification.forward`,  you can safely ignore this message.\n",
+      "/mounts/work/akoksal/anaconda3/envs/lmbias/lib/python3.9/site-packages/transformers/optimization.py:306: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use the PyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning\n",
+      "  warnings.warn(\n",
+      "***** Running training *****\n",
+      "  Num examples = 799\n",
+      "  Num Epochs = 5\n",
+      "  Instantaneous batch size per device = 16\n",
+      "  Total train batch size (w. parallel, distributed & accumulation) = 16\n",
+      "  Gradient Accumulation steps = 1\n",
+      "  Total optimization steps = 250\n",
+      "  Number of trainable parameters = 183773977\n",
+      "You're using a BertTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.\n"
+     ]
+    },
+    {
+     "data": {
+      "text/html": [
+       "\n",
+       "    <div>\n",
+       "      \n",
+       "      <progress value='250' max='250' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
+       "      [250/250 01:12, Epoch 5/5]\n",
+       "    </div>\n",
+       "    <table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       " <tr style=\"text-align: left;\">\n",
+       "      <th>Epoch</th>\n",
+       "      <th>Training Loss</th>\n",
+       "      <th>Validation Loss</th>\n",
+       "      <th>Overall Precision</th>\n",
+       "      <th>Overall Recall</th>\n",
+       "      <th>Overall F1</th>\n",
+       "      <th>Overall Accuracy</th>\n",
+       "      <th>Bina F1</th>\n",
+       "      <th>Bina Recall</th>\n",
+       "      <th>Bina Precision</th>\n",
+       "      <th>Bina Support</th>\n",
+       "      <th>Bulvar F1</th>\n",
+       "      <th>Bulvar Recall</th>\n",
+       "      <th>Bulvar Precision</th>\n",
+       "      <th>Bulvar Support</th>\n",
+       "      <th>Cadde F1</th>\n",
+       "      <th>Cadde Recall</th>\n",
+       "      <th>Cadde Precision</th>\n",
+       "      <th>Cadde Support</th>\n",
+       "      <th>Diskapino F1</th>\n",
+       "      <th>Diskapino Recall</th>\n",
+       "      <th>Diskapino Precision</th>\n",
+       "      <th>Diskapino Support</th>\n",
+       "      <th>Ilce F1</th>\n",
+       "      <th>Ilce Recall</th>\n",
+       "      <th>Ilce Precision</th>\n",
+       "      <th>Ilce Support</th>\n",
+       "      <th>Isim F1</th>\n",
+       "      <th>Isim Recall</th>\n",
+       "      <th>Isim Precision</th>\n",
+       "      <th>Isim Support</th>\n",
+       "      <th>Mahalle F1</th>\n",
+       "      <th>Mahalle Recall</th>\n",
+       "      <th>Mahalle Precision</th>\n",
+       "      <th>Mahalle Support</th>\n",
+       "      <th>Sehir F1</th>\n",
+       "      <th>Sehir Recall</th>\n",
+       "      <th>Sehir Precision</th>\n",
+       "      <th>Sehir Support</th>\n",
+       "      <th>Site F1</th>\n",
+       "      <th>Site Recall</th>\n",
+       "      <th>Site Precision</th>\n",
+       "      <th>Site Support</th>\n",
+       "      <th>Sokak F1</th>\n",
+       "      <th>Sokak Recall</th>\n",
+       "      <th>Sokak Precision</th>\n",
+       "      <th>Sokak Support</th>\n",
+       "      <th>Soyisim F1</th>\n",
+       "      <th>Soyisim Recall</th>\n",
+       "      <th>Soyisim Precision</th>\n",
+       "      <th>Soyisim Support</th>\n",
+       "      <th>Telefonno F1</th>\n",
+       "      <th>Telefonno Recall</th>\n",
+       "      <th>Telefonno Precision</th>\n",
+       "      <th>Telefonno Support</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <td>1</td>\n",
+       "      <td>1.349500</td>\n",
+       "      <td>0.357321</td>\n",
+       "      <td>0.783270</td>\n",
+       "      <td>0.828974</td>\n",
+       "      <td>0.805474</td>\n",
+       "      <td>0.908936</td>\n",
+       "      <td>0.600000</td>\n",
+       "      <td>0.705882</td>\n",
+       "      <td>0.521739</td>\n",
+       "      <td>34</td>\n",
+       "      <td>0.000000</td>\n",
+       "      <td>0.000000</td>\n",
+       "      <td>0.000000</td>\n",
+       "      <td>5</td>\n",
+       "      <td>0.588235</td>\n",
+       "      <td>0.833333</td>\n",
+       "      <td>0.454545</td>\n",
+       "      <td>24</td>\n",
+       "      <td>0.769231</td>\n",
+       "      <td>0.892857</td>\n",
+       "      <td>0.675676</td>\n",
+       "      <td>28</td>\n",
+       "      <td>0.830508</td>\n",
+       "      <td>0.816667</td>\n",
+       "      <td>0.844828</td>\n",
+       "      <td>60</td>\n",
+       "      <td>0.888889</td>\n",
+       "      <td>0.926829</td>\n",
+       "      <td>0.853933</td>\n",
+       "      <td>82</td>\n",
+       "      <td>0.750000</td>\n",
+       "      <td>0.792453</td>\n",
+       "      <td>0.711864</td>\n",
+       "      <td>53</td>\n",
+       "      <td>0.867133</td>\n",
+       "      <td>0.861111</td>\n",
+       "      <td>0.873239</td>\n",
+       "      <td>72</td>\n",
+       "      <td>0.000000</td>\n",
+       "      <td>0.000000</td>\n",
+       "      <td>0.000000</td>\n",
+       "      <td>6</td>\n",
+       "      <td>0.750000</td>\n",
+       "      <td>0.620690</td>\n",
+       "      <td>0.947368</td>\n",
+       "      <td>29</td>\n",
+       "      <td>0.900000</td>\n",
+       "      <td>0.887324</td>\n",
+       "      <td>0.913043</td>\n",
+       "      <td>71</td>\n",
+       "      <td>0.985075</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.970588</td>\n",
+       "      <td>33</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <td>2</td>\n",
+       "      <td>0.264700</td>\n",
+       "      <td>0.220467</td>\n",
+       "      <td>0.885149</td>\n",
+       "      <td>0.899396</td>\n",
+       "      <td>0.892216</td>\n",
+       "      <td>0.944792</td>\n",
+       "      <td>0.782609</td>\n",
+       "      <td>0.794118</td>\n",
+       "      <td>0.771429</td>\n",
+       "      <td>34</td>\n",
+       "      <td>0.666667</td>\n",
+       "      <td>0.800000</td>\n",
+       "      <td>0.571429</td>\n",
+       "      <td>5</td>\n",
+       "      <td>0.875000</td>\n",
+       "      <td>0.875000</td>\n",
+       "      <td>0.875000</td>\n",
+       "      <td>24</td>\n",
+       "      <td>0.862069</td>\n",
+       "      <td>0.892857</td>\n",
+       "      <td>0.833333</td>\n",
+       "      <td>28</td>\n",
+       "      <td>0.894309</td>\n",
+       "      <td>0.916667</td>\n",
+       "      <td>0.873016</td>\n",
+       "      <td>60</td>\n",
+       "      <td>0.884848</td>\n",
+       "      <td>0.890244</td>\n",
+       "      <td>0.879518</td>\n",
+       "      <td>82</td>\n",
+       "      <td>0.897196</td>\n",
+       "      <td>0.905660</td>\n",
+       "      <td>0.888889</td>\n",
+       "      <td>53</td>\n",
+       "      <td>0.915493</td>\n",
+       "      <td>0.902778</td>\n",
+       "      <td>0.928571</td>\n",
+       "      <td>72</td>\n",
+       "      <td>0.181818</td>\n",
+       "      <td>0.166667</td>\n",
+       "      <td>0.200000</td>\n",
+       "      <td>6</td>\n",
+       "      <td>0.949153</td>\n",
+       "      <td>0.965517</td>\n",
+       "      <td>0.933333</td>\n",
+       "      <td>29</td>\n",
+       "      <td>0.950355</td>\n",
+       "      <td>0.943662</td>\n",
+       "      <td>0.957143</td>\n",
+       "      <td>71</td>\n",
+       "      <td>0.985075</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.970588</td>\n",
+       "      <td>33</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <td>3</td>\n",
+       "      <td>0.158700</td>\n",
+       "      <td>0.219565</td>\n",
+       "      <td>0.876768</td>\n",
+       "      <td>0.873239</td>\n",
+       "      <td>0.875000</td>\n",
+       "      <td>0.940808</td>\n",
+       "      <td>0.805556</td>\n",
+       "      <td>0.852941</td>\n",
+       "      <td>0.763158</td>\n",
+       "      <td>34</td>\n",
+       "      <td>0.666667</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.500000</td>\n",
+       "      <td>5</td>\n",
+       "      <td>0.880000</td>\n",
+       "      <td>0.916667</td>\n",
+       "      <td>0.846154</td>\n",
+       "      <td>24</td>\n",
+       "      <td>0.827586</td>\n",
+       "      <td>0.857143</td>\n",
+       "      <td>0.800000</td>\n",
+       "      <td>28</td>\n",
+       "      <td>0.881356</td>\n",
+       "      <td>0.866667</td>\n",
+       "      <td>0.896552</td>\n",
+       "      <td>60</td>\n",
+       "      <td>0.822785</td>\n",
+       "      <td>0.792683</td>\n",
+       "      <td>0.855263</td>\n",
+       "      <td>82</td>\n",
+       "      <td>0.886792</td>\n",
+       "      <td>0.886792</td>\n",
+       "      <td>0.886792</td>\n",
+       "      <td>53</td>\n",
+       "      <td>0.892086</td>\n",
+       "      <td>0.861111</td>\n",
+       "      <td>0.925373</td>\n",
+       "      <td>72</td>\n",
+       "      <td>0.400000</td>\n",
+       "      <td>0.333333</td>\n",
+       "      <td>0.500000</td>\n",
+       "      <td>6</td>\n",
+       "      <td>0.881356</td>\n",
+       "      <td>0.896552</td>\n",
+       "      <td>0.866667</td>\n",
+       "      <td>29</td>\n",
+       "      <td>0.957143</td>\n",
+       "      <td>0.943662</td>\n",
+       "      <td>0.971014</td>\n",
+       "      <td>71</td>\n",
+       "      <td>0.985075</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.970588</td>\n",
+       "      <td>33</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <td>4</td>\n",
+       "      <td>0.115000</td>\n",
+       "      <td>0.215329</td>\n",
+       "      <td>0.897541</td>\n",
+       "      <td>0.881288</td>\n",
+       "      <td>0.889340</td>\n",
+       "      <td>0.946500</td>\n",
+       "      <td>0.857143</td>\n",
+       "      <td>0.882353</td>\n",
+       "      <td>0.833333</td>\n",
+       "      <td>34</td>\n",
+       "      <td>0.909091</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.833333</td>\n",
+       "      <td>5</td>\n",
+       "      <td>0.897959</td>\n",
+       "      <td>0.916667</td>\n",
+       "      <td>0.880000</td>\n",
+       "      <td>24</td>\n",
+       "      <td>0.862069</td>\n",
+       "      <td>0.892857</td>\n",
+       "      <td>0.833333</td>\n",
+       "      <td>28</td>\n",
+       "      <td>0.881356</td>\n",
+       "      <td>0.866667</td>\n",
+       "      <td>0.896552</td>\n",
+       "      <td>60</td>\n",
+       "      <td>0.810127</td>\n",
+       "      <td>0.780488</td>\n",
+       "      <td>0.842105</td>\n",
+       "      <td>82</td>\n",
+       "      <td>0.886792</td>\n",
+       "      <td>0.886792</td>\n",
+       "      <td>0.886792</td>\n",
+       "      <td>53</td>\n",
+       "      <td>0.890511</td>\n",
+       "      <td>0.847222</td>\n",
+       "      <td>0.938462</td>\n",
+       "      <td>72</td>\n",
+       "      <td>0.727273</td>\n",
+       "      <td>0.666667</td>\n",
+       "      <td>0.800000</td>\n",
+       "      <td>6</td>\n",
+       "      <td>0.950820</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.906250</td>\n",
+       "      <td>29</td>\n",
+       "      <td>0.949640</td>\n",
+       "      <td>0.929577</td>\n",
+       "      <td>0.970588</td>\n",
+       "      <td>71</td>\n",
+       "      <td>0.985075</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.970588</td>\n",
+       "      <td>33</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <td>5</td>\n",
+       "      <td>0.093800</td>\n",
+       "      <td>0.231558</td>\n",
+       "      <td>0.895492</td>\n",
+       "      <td>0.879276</td>\n",
+       "      <td>0.887310</td>\n",
+       "      <td>0.945361</td>\n",
+       "      <td>0.833333</td>\n",
+       "      <td>0.882353</td>\n",
+       "      <td>0.789474</td>\n",
+       "      <td>34</td>\n",
+       "      <td>0.909091</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.833333</td>\n",
+       "      <td>5</td>\n",
+       "      <td>0.880000</td>\n",
+       "      <td>0.916667</td>\n",
+       "      <td>0.846154</td>\n",
+       "      <td>24</td>\n",
+       "      <td>0.813559</td>\n",
+       "      <td>0.857143</td>\n",
+       "      <td>0.774194</td>\n",
+       "      <td>28</td>\n",
+       "      <td>0.888889</td>\n",
+       "      <td>0.866667</td>\n",
+       "      <td>0.912281</td>\n",
+       "      <td>60</td>\n",
+       "      <td>0.833333</td>\n",
+       "      <td>0.792683</td>\n",
+       "      <td>0.878378</td>\n",
+       "      <td>82</td>\n",
+       "      <td>0.895238</td>\n",
+       "      <td>0.886792</td>\n",
+       "      <td>0.903846</td>\n",
+       "      <td>53</td>\n",
+       "      <td>0.898551</td>\n",
+       "      <td>0.861111</td>\n",
+       "      <td>0.939394</td>\n",
+       "      <td>72</td>\n",
+       "      <td>0.727273</td>\n",
+       "      <td>0.666667</td>\n",
+       "      <td>0.800000</td>\n",
+       "      <td>6</td>\n",
+       "      <td>0.881356</td>\n",
+       "      <td>0.896552</td>\n",
+       "      <td>0.866667</td>\n",
+       "      <td>29</td>\n",
+       "      <td>0.957143</td>\n",
+       "      <td>0.943662</td>\n",
+       "      <td>0.971014</td>\n",
+       "      <td>71</td>\n",
+       "      <td>0.985075</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>0.970588</td>\n",
+       "      <td>33</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table><p>"
+      ],
+      "text/plain": [
+       "<IPython.core.display.HTML object>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "The following columns in the evaluation set don't have a corresponding argument in `BertForTokenClassification.forward` and have been ignored: tokens, ids, ner_tags. If tokens, ids, ner_tags are not expected by `BertForTokenClassification.forward`,  you can safely ignore this message.\n",
+      "***** Running Evaluation *****\n",
+      "  Num examples = 58\n",
+      "  Batch size = 32\n",
+      "/mounts/work/akoksal/anaconda3/envs/lmbias/lib/python3.9/site-packages/seqeval/metrics/v1.py:57: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
+      "  _warn_prf(average, modifier, msg_start, len(result))\n",
+      "Saving model checkpoint to /mounts/work/akoksal/earthquake_ner_models/checkpoint-50\n",
+      "Configuration saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-50/config.json\n",
+      "Model weights saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-50/pytorch_model.bin\n",
+      "tokenizer config file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-50/tokenizer_config.json\n",
+      "Special tokens file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-50/special_tokens_map.json\n",
+      "The following columns in the evaluation set don't have a corresponding argument in `BertForTokenClassification.forward` and have been ignored: tokens, ids, ner_tags. If tokens, ids, ner_tags are not expected by `BertForTokenClassification.forward`,  you can safely ignore this message.\n",
+      "***** Running Evaluation *****\n",
+      "  Num examples = 58\n",
+      "  Batch size = 32\n",
+      "Saving model checkpoint to /mounts/work/akoksal/earthquake_ner_models/checkpoint-100\n",
+      "Configuration saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-100/config.json\n",
+      "Model weights saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-100/pytorch_model.bin\n",
+      "tokenizer config file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-100/tokenizer_config.json\n",
+      "Special tokens file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-100/special_tokens_map.json\n",
+      "The following columns in the evaluation set don't have a corresponding argument in `BertForTokenClassification.forward` and have been ignored: tokens, ids, ner_tags. If tokens, ids, ner_tags are not expected by `BertForTokenClassification.forward`,  you can safely ignore this message.\n",
+      "***** Running Evaluation *****\n",
+      "  Num examples = 58\n",
+      "  Batch size = 32\n",
+      "Saving model checkpoint to /mounts/work/akoksal/earthquake_ner_models/checkpoint-150\n",
+      "Configuration saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-150/config.json\n",
+      "Model weights saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-150/pytorch_model.bin\n",
+      "tokenizer config file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-150/tokenizer_config.json\n",
+      "Special tokens file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-150/special_tokens_map.json\n",
+      "The following columns in the evaluation set don't have a corresponding argument in `BertForTokenClassification.forward` and have been ignored: tokens, ids, ner_tags. If tokens, ids, ner_tags are not expected by `BertForTokenClassification.forward`,  you can safely ignore this message.\n",
+      "***** Running Evaluation *****\n",
+      "  Num examples = 58\n",
+      "  Batch size = 32\n",
+      "Saving model checkpoint to /mounts/work/akoksal/earthquake_ner_models/checkpoint-200\n",
+      "Configuration saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-200/config.json\n",
+      "Model weights saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-200/pytorch_model.bin\n",
+      "tokenizer config file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-200/tokenizer_config.json\n",
+      "Special tokens file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-200/special_tokens_map.json\n",
+      "Deleting older checkpoint [/mounts/work/akoksal/earthquake_ner_models/checkpoint-50] due to args.save_total_limit\n",
+      "The following columns in the evaluation set don't have a corresponding argument in `BertForTokenClassification.forward` and have been ignored: tokens, ids, ner_tags. If tokens, ids, ner_tags are not expected by `BertForTokenClassification.forward`,  you can safely ignore this message.\n",
+      "***** Running Evaluation *****\n",
+      "  Num examples = 58\n",
+      "  Batch size = 32\n",
+      "Saving model checkpoint to /mounts/work/akoksal/earthquake_ner_models/checkpoint-250\n",
+      "Configuration saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-250/config.json\n",
+      "Model weights saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-250/pytorch_model.bin\n",
+      "tokenizer config file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-250/tokenizer_config.json\n",
+      "Special tokens file saved in /mounts/work/akoksal/earthquake_ner_models/checkpoint-250/special_tokens_map.json\n",
+      "Deleting older checkpoint [/mounts/work/akoksal/earthquake_ner_models/checkpoint-100] due to args.save_total_limit\n",
+      "\n",
+      "\n",
+      "Training completed. Do not forget to share your model on huggingface.co/models =)\n",
+      "\n",
+      "\n",
+      "Loading best model from /mounts/work/akoksal/earthquake_ner_models/checkpoint-200 (score: 0.21532948315143585).\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "TrainOutput(global_step=250, training_loss=0.3963502960205078, metrics={'train_runtime': 73.0701, 'train_samples_per_second': 54.674, 'train_steps_per_second': 3.421, 'total_flos': 129863927953500.0, 'train_loss': 0.3963502960205078, 'epoch': 5.0})"
+      ]
+     },
+     "execution_count": 15,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "trainer.train()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "id": "4427c32d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "The following columns in the evaluation set don't have a corresponding argument in `BertForTokenClassification.forward` and have been ignored: tokens, ids, ner_tags. If tokens, ids, ner_tags are not expected by `BertForTokenClassification.forward`,  you can safely ignore this message.\n",
+      "***** Running Evaluation *****\n",
+      "  Num examples = 129\n",
+      "  Batch size = 32\n"
+     ]
+    },
+    {
+     "data": {
+      "text/html": [
+       "\n",
+       "    <div>\n",
+       "      \n",
+       "      <progress value='5' max='5' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
+       "      [5/5 00:00]\n",
+       "    </div>\n",
+       "    "
+      ],
+      "text/plain": [
+       "<IPython.core.display.HTML object>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "results = trainer.evaluate(tokenized_dataset[\"test\"])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 24,
+   "id": "aabbb977",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "{'eval_loss': 0.24822480976581573,\n",
+       " 'eval_overall_precision': 0.8442211055276382,\n",
+       " 'eval_overall_recall': 0.877742946708464,\n",
+       " 'eval_overall_f1': 0.860655737704918,\n",
+       " 'eval_overall_accuracy': 0.9401853411962932,\n",
+       " 'eval_bina_f1': 0.7000000000000001,\n",
+       " 'eval_bina_recall': 0.7424242424242424,\n",
+       " 'eval_bina_precision': 0.6621621621621622,\n",
+       " 'eval_bina_support': 66,\n",
+       " 'eval_bulvar_f1': 0.9230769230769231,\n",
+       " 'eval_bulvar_recall': 0.9230769230769231,\n",
+       " 'eval_bulvar_precision': 0.9230769230769231,\n",
+       " 'eval_bulvar_support': 13,\n",
+       " 'eval_cadde_f1': 0.8067226890756302,\n",
+       " 'eval_cadde_recall': 0.8421052631578947,\n",
+       " 'eval_cadde_precision': 0.7741935483870968,\n",
+       " 'eval_cadde_support': 57,\n",
+       " 'eval_diskapino_f1': 0.7083333333333334,\n",
+       " 'eval_diskapino_recall': 0.7285714285714285,\n",
+       " 'eval_diskapino_precision': 0.6891891891891891,\n",
+       " 'eval_diskapino_support': 70,\n",
+       " 'eval_ilce_f1': 0.9218106995884773,\n",
+       " 'eval_ilce_recall': 0.9572649572649573,\n",
+       " 'eval_ilce_precision': 0.8888888888888888,\n",
+       " 'eval_ilce_support': 117,\n",
+       " 'eval_isim_f1': 0.8793103448275862,\n",
+       " 'eval_isim_recall': 0.9026548672566371,\n",
+       " 'eval_isim_precision': 0.8571428571428571,\n",
+       " 'eval_isim_support': 113,\n",
+       " 'eval_mahalle_f1': 0.7903225806451613,\n",
+       " 'eval_mahalle_recall': 0.8166666666666667,\n",
+       " 'eval_mahalle_precision': 0.765625,\n",
+       " 'eval_mahalle_support': 120,\n",
+       " 'eval_sehir_f1': 0.9724137931034483,\n",
+       " 'eval_sehir_recall': 0.9657534246575342,\n",
+       " 'eval_sehir_precision': 0.9791666666666666,\n",
+       " 'eval_sehir_support': 146,\n",
+       " 'eval_site_f1': 0.6875000000000001,\n",
+       " 'eval_site_recall': 0.6111111111111112,\n",
+       " 'eval_site_precision': 0.7857142857142857,\n",
+       " 'eval_site_support': 18,\n",
+       " 'eval_sokak_f1': 0.7301587301587302,\n",
+       " 'eval_sokak_recall': 0.7419354838709677,\n",
+       " 'eval_sokak_precision': 0.71875,\n",
+       " 'eval_sokak_support': 62,\n",
+       " 'eval_soyisim_f1': 0.9441624365482234,\n",
+       " 'eval_soyisim_recall': 0.9489795918367347,\n",
+       " 'eval_soyisim_precision': 0.9393939393939394,\n",
+       " 'eval_soyisim_support': 98,\n",
+       " 'eval_telefonno_f1': 0.9935483870967742,\n",
+       " 'eval_telefonno_recall': 1.0,\n",
+       " 'eval_telefonno_precision': 0.9871794871794872,\n",
+       " 'eval_telefonno_support': 77,\n",
+       " 'eval_runtime': 0.3493,\n",
+       " 'eval_samples_per_second': 369.308,\n",
+       " 'eval_steps_per_second': 14.314,\n",
+       " 'epoch': 5.0}"
+      ]
+     },
+     "execution_count": 24,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "results"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "id": "922a7237",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>support</th>\n",
+       "      <th>precision</th>\n",
+       "      <th>recall</th>\n",
+       "      <th>f1</th>\n",
+       "      <th>accuracy</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>overall</th>\n",
+       "      <td>957</td>\n",
+       "      <td>0.84</td>\n",
+       "      <td>0.88</td>\n",
+       "      <td>0.86</td>\n",
+       "      <td>0.94</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>bina</th>\n",
+       "      <td>66</td>\n",
+       "      <td>0.66</td>\n",
+       "      <td>0.74</td>\n",
+       "      <td>0.70</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>bulvar</th>\n",
+       "      <td>13</td>\n",
+       "      <td>0.92</td>\n",
+       "      <td>0.92</td>\n",
+       "      <td>0.92</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>cadde</th>\n",
+       "      <td>57</td>\n",
+       "      <td>0.77</td>\n",
+       "      <td>0.84</td>\n",
+       "      <td>0.81</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>diskapino</th>\n",
+       "      <td>70</td>\n",
+       "      <td>0.69</td>\n",
+       "      <td>0.73</td>\n",
+       "      <td>0.71</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>ilce</th>\n",
+       "      <td>117</td>\n",
+       "      <td>0.89</td>\n",
+       "      <td>0.96</td>\n",
+       "      <td>0.92</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>isim</th>\n",
+       "      <td>113</td>\n",
+       "      <td>0.86</td>\n",
+       "      <td>0.90</td>\n",
+       "      <td>0.88</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>mahalle</th>\n",
+       "      <td>120</td>\n",
+       "      <td>0.77</td>\n",
+       "      <td>0.82</td>\n",
+       "      <td>0.79</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>sehir</th>\n",
+       "      <td>146</td>\n",
+       "      <td>0.98</td>\n",
+       "      <td>0.97</td>\n",
+       "      <td>0.97</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>site</th>\n",
+       "      <td>18</td>\n",
+       "      <td>0.79</td>\n",
+       "      <td>0.61</td>\n",
+       "      <td>0.69</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>sokak</th>\n",
+       "      <td>62</td>\n",
+       "      <td>0.72</td>\n",
+       "      <td>0.74</td>\n",
+       "      <td>0.73</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>soyisim</th>\n",
+       "      <td>98</td>\n",
+       "      <td>0.94</td>\n",
+       "      <td>0.95</td>\n",
+       "      <td>0.94</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>telefonno</th>\n",
+       "      <td>77</td>\n",
+       "      <td>0.99</td>\n",
+       "      <td>1.00</td>\n",
+       "      <td>0.99</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "           support  precision  recall    f1  accuracy\n",
+       "overall        957       0.84    0.88  0.86      0.94\n",
+       "bina            66       0.66    0.74  0.70       NaN\n",
+       "bulvar          13       0.92    0.92  0.92       NaN\n",
+       "cadde           57       0.77    0.84  0.81       NaN\n",
+       "diskapino       70       0.69    0.73  0.71       NaN\n",
+       "ilce           117       0.89    0.96  0.92       NaN\n",
+       "isim           113       0.86    0.90  0.88       NaN\n",
+       "mahalle        120       0.77    0.82  0.79       NaN\n",
+       "sehir          146       0.98    0.97  0.97       NaN\n",
+       "site            18       0.79    0.61  0.69       NaN\n",
+       "sokak           62       0.72    0.74  0.73       NaN\n",
+       "soyisim         98       0.94    0.95  0.94       NaN\n",
+       "telefonno       77       0.99    1.00  0.99       NaN"
+      ]
+     },
+     "execution_count": 18,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "structured_results = defaultdict(dict)\n",
+    "structured_results[\"overall\"][\"support\"]=0\n",
+    "for x, y in results.items():\n",
+    "    if len(x.split(\"_\"))==3:\n",
+    "        structured_results[x.split(\"_\")[1]][x.split(\"_\")[2]] = y\n",
+    "        if x.split(\"_\")[2]==\"support\":\n",
+    "            structured_results[\"overall\"][\"support\"]+=y\n",
+    "results_pd = pd.DataFrame(structured_results).T\n",
+    "results_pd.support = results_pd.support.astype(int)\n",
+    "results_pd.round(2)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "3c3de283",
+   "metadata": {},
+   "source": [
+    "## Predictions"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "id": "ed165edb",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from transformers import pipeline\n",
+    "nlp = pipeline(\"ner\", model=model.to(device), tokenizer=tokenizer, aggregation_strategy=\"first\", device=0 if device==\"cuda\" else -1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 20,
+   "id": "0e350503",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Source: https://www.thepythoncode.com/article/named-entity-recognition-using-transformers-and-spacy\n",
+    "def get_entities_html(text, ner_result, title=None):\n",
+    "    \"\"\"Visualize NER with the help of SpaCy\"\"\"\n",
+    "    ents = []\n",
+    "    for ent in ner_result:\n",
+    "        e = {}\n",
+    "        # add the start and end positions of the entity\n",
+    "        e[\"start\"] = ent[\"start\"]\n",
+    "        e[\"end\"] = ent[\"end\"]\n",
+    "        # add the score if you want in the label\n",
+    "        # e[\"label\"] = f\"{ent[\"entity\"]}-{ent['score']:.2f}\"\n",
+    "        e[\"label\"] = ent[\"entity_group\"]\n",
+    "        if ents and -1 <= ent[\"start\"] - ents[-1][\"end\"] <= 1 and ents[-1][\"label\"] == e[\"label\"]:\n",
+    "            # if the current entity is shared with previous entity\n",
+    "            # simply extend the entity end position instead of adding a new one\n",
+    "            ents[-1][\"end\"] = e[\"end\"]\n",
+    "            continue\n",
+    "        ents.append(e)\n",
+    "      # construct data required for displacy.render() method\n",
+    "    render_data = [\n",
+    "    {\n",
+    "      \"text\": text,\n",
+    "      \"ents\": ents,\n",
+    "      \"title\": title,\n",
+    "    }\n",
+    "    ]\n",
+    "    spacy.displacy.render(render_data, style=\"ent\", manual=True, jupyter=True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 21,
+   "id": "f98a6902",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<span class=\"tex2jax_ignore\"><div class=\"entities\" style=\"line-height: 2.5; direction: ltr\">Lütfen yardım \n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    Akevler\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">mahalle</span>\n",
+       "</mark>\n",
+       " mahallesi \n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    Rüzgar\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">sokak</span>\n",
+       "</mark>\n",
+       " sokak \n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    Tuncay\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">bina</span>\n",
+       "</mark>\n",
+       " apartmanı zemin kat \n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    Antakya\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ilce</span>\n",
+       "</mark>\n",
+       " akrabalarım göçük altında #hatay #Afad</div></span>"
+      ],
+      "text/plain": [
+       "<IPython.core.display.HTML object>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "sentence = \"\"\"Lütfen yardım Akevler mahallesi Rüzgar sokak Tuncay apartmanı zemin kat Antakya akrabalarım göçük altında #hatay #Afad\"\"\"\n",
+    "\n",
+    "get_entities_html(sentence, nlp(sentence))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "id": "80b823ff",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<span class=\"tex2jax_ignore\"><div class=\"entities\" style=\"line-height: 2.5; direction: ltr\">\n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    Kahramanmaraş\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">sehir</span>\n",
+       "</mark>\n",
+       " \n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    merkez\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ilce</span>\n",
+       "</mark>\n",
+       " \n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    Şazibey\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">mahalle</span>\n",
+       "</mark>\n",
+       " Mahallesi \n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    Ebrar\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">site</span>\n",
+       "</mark>\n",
+       " Sitesi \n",
+       "<mark class=\"entity\" style=\"background: #ddd; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
+       "    Z\n",
+       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">bina</span>\n",
+       "</mark>\n",
+       " blok arka tarafı için acil en az 150 tonluk vinç lazım lütfen paylaşır mısınız</div></span>"
+      ],
+      "text/plain": [
+       "<IPython.core.display.HTML object>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "sentence = \" \".join(dataset[\"train\"][433][\"tokens\"])\n",
+    "get_entities_html(sentence, nlp(sentence))"
+   ]
+  }
+ ],
+ "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": "3.9.12"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}