inclusion_transformers_2024update.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2025-01-01 15:19:18.297785: E tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:9342] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
      "2025-01-01 15:19:18.297838: E tensorflow/compiler/xla/stream_executor/cuda/cuda_fft.cc:609] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
      "2025-01-01 15:19:18.297866: E tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:1518] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
      "2025-01-01 15:19:18.303814: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
      "To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2025-01-01 15:19:19.124033: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import pickle\n",
    "import scipy as sp\n",
    "import shap\n",
    "import torch\n",
    "import os\n",
    "\n",
    "from datasets import Dataset, DatasetDict\n",
    "from matplotlib import pyplot as plt\n",
    "from sklearn.metrics import accuracy_score, f1_score, precision_recall_curve, PrecisionRecallDisplay, precision_score, recall_score\n",
    "from sklearn.model_selection import train_test_split, KFold\n",
    "from tqdm.notebook import trange\n",
    "from ray import tune\n",
    "from transformers import AutoTokenizer, AutoModelForSequenceClassification, TextClassificationPipeline, TrainingArguments, Trainer\n",
    "\n",
    "os.environ[\"TOKENIZERS_PARALLELISM\"] = \"false\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "### YOUR PATH HERE\n",
    "\n",
    "#For execution in a devcontainer on a Linux server with Docker\n",
    "filepath = '/workspaces/impactpharmacie'\n",
    "\n",
    "#For execution in a devcontainer on a Windows PC running WSL with Docker\n",
    "#filepath = '.'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Import the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of samples in all df: 12024, Number of samples in test df: 2400, Number of samples in train_val df: 9624\n"
     ]
    }
   ],
   "source": [
    "df = pd.read_csv(filepath + '/data/static_oct2024_update/ratings.csv')[1:].fillna('')\n",
    "df['labels'] = df.apply(lambda x: x['rating_consensus'] if x['rating_consensus'] != '' else x['rating1'] if ((x['rating1'] == x['rating2']) and (x['consensus_reason'] == '') and (x['rating1'] != '')) else 'error', axis=1)\n",
    "test_df = df[-2400:]\n",
    "train_val_df = df[:-2400]\n",
    "print('Number of samples in all df: {}, Number of samples in test df: {}, Number of samples in train_val df: {}'.format(len(df),len(test_df), len(train_val_df)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Length of train set: 7699\n",
      "Length of val set: 1925\n",
      "Number of included in train set: 488.0 for ratio: 0.063\n",
      "Number of included in val set: 101.0 for ratio: 0.052\n"
     ]
    }
   ],
   "source": [
    "train_df, val_df = train_test_split (train_val_df, test_size=0.2)\n",
    "\n",
    "print('Length of train set: {}'.format(len(train_df)))\n",
    "print('Length of val set: {}'.format(len(val_df)))\n",
    "print('Number of included in train set: {} for ratio: {:.3f}'.format(train_df['labels'].sum(), train_df['labels'].sum()/len(train_df)))\n",
    "print('Number of included in val set: {} for ratio: {:.3f}'.format(val_df['labels'].sum(), val_df['labels'].sum()/len(val_df)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_columns_to_keep = ['text', 'labels']\n",
    "train_val_df = train_val_df[df_columns_to_keep]\n",
    "train_df = train_df[df_columns_to_keep]\n",
    "val_df = val_df[df_columns_to_keep]\n",
    "test_df = test_df[df_columns_to_keep]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_val_ds = Dataset.from_pandas(train_val_df)\n",
    "train_ds = Dataset.from_pandas(train_df)\n",
    "val_ds = Dataset.from_pandas(val_df)\n",
    "test_ds = Dataset.from_pandas(test_df)\n",
    "\n",
    "ds = DatasetDict()\n",
    "ds['train_val'] = train_val_ds\n",
    "ds['train'] = train_ds\n",
    "ds['val'] = val_ds\n",
    "ds['test'] = test_ds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ds.cleanup_cache_files()\n",
    "test_ds.cleanup_cache_files()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a55ede9f78a54dee93e35ed179dc4c32",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/9624 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "99486db2233f4b4bb5c21a381c3e64f0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/9624 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "433b6f107a9449b582acdb31135827ff",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/7699 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "08bb03bfd8cc4142bf5f5d2e8fc63e89",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/7699 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2b2e3de81725449e9d7ae93016c84757",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/1925 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6cfaadc9b983443c81251369224659bb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/1925 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8bbfc64bb6f04700a2bea1658e52ad15",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/2400 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a3e2566d9a7948dea8fe64e5b884cec1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/2400 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ds = ds.class_encode_column('labels')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Helpful functions for the models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_single_label_metrics(eval_pred):\n",
    "    logits, labels = eval_pred\n",
    "    predictions = np.argmax(logits, axis=1)\n",
    "    metrics = {'accuracy':accuracy_score(labels, predictions), 'f1':f1_score(labels, predictions), 'precision':precision_score(labels, predictions), 'recall':recall_score(labels, predictions)}\n",
    "    return metrics\n",
    "\n",
    "def tokenizer_function(samples):\n",
    "    return tokenizer(samples['text'], padding=\"max_length\", truncation=True, max_length=512)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluate different models\n",
    "\n",
    "List of models to evaluate based on exploration of models available in Huggingface and cited in https://doi.org/10.1016/j.compbiomed.2024.108189 ( A comprehensive evaluation of large Language models on benchmark biomedical text processing tasks )\n",
    "\n",
    "Planned strategy, list of models to be evaluated:\n",
    "1. BiomedBERT base-abstract-fulltext // microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext\n",
    "2. BiomedBERT base-abstract // microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract\n",
    "3. BiomedBERT large-abstract // microsoft/BiomedNLP-BiomedBERT-large-uncased-abstract\n",
    "4. BioBERT base-1.2 // dmis-lab/biobert-base-cased-v1.2\n",
    "5. BioBERT large-1.1 // dmis-lab/biobert-large-cased-v1.1\n",
    "6. BioBART v2-base // GanjinZero/biobart-v2-base\n",
    "7. BioBART v2-large // GanjinZero/biobart-v2-large\n",
    "8. BioELECTRA base-discriminator-pubmed // kamalkraj/bioelectra-base-discriminator-pubmed\n",
    "9. BioELECTRA base-discriminator-pubmed-pmc // kamalkraj/bioelectra-base-discriminator-pubmed-pmc\n",
    "10. BioELECTRA base-discriminator-pubmed-pmc-lt // kamalkraj/bioelectra-base-discriminator-pubmed-pmc-lt\n",
    "11. BioGPT // microsoft/biogpt\n",
    "12. BioGPT-large // microsoft/BioGPT-Large\n",
    "13. BioGPT-Large-PubMedQA // microsoft/BioGPT-Large-PubMedQA\n",
    "\n",
    "(BioBART models failed because or a tensor dimension error - seems to be incompatible with sequence classification models)\n",
    "(BioELECTRA models can't even initialize - seem to need local training)\n",
    "(BioGPT models are too large for the GPU of our current server - model training starts but then fails because of OOM error on the GPU - not evaluated)\n",
    "\n",
    "Cell outputs removed from the notebook because too large."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "UsageError: Line magic function `%tensorboard` not found.\n"
     ]
    }
   ],
   "source": [
    "%tensorboard"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run1- Do a general hyperparameter search on the models to find candidates with reasonable performance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BiomedBERT base-abstract-fulltext"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = 'microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext'\n",
    "model_short_name = 'BiomedBERT-base-abstract-fulltext'\n",
    "run_number = '1'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.choice([5e-5, 2e-5, 1e-5, 8e-6, 5e-6, 2e-6, 1e-6]),\n",
    "    'num_train_epochs':tune.choice([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':8\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=40, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BiomedBERT base-abstract"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = 'microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract'\n",
    "model_short_name = 'BiomedBERT-base-abstract'\n",
    "run_number = '1'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.choice([5e-5, 2e-5, 1e-5, 8e-6, 5e-6, 2e-6, 1e-6]),\n",
    "    'num_train_epochs':tune.choice([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':8\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=40, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BiomedBERT large-abstract"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = 'microsoft/BiomedNLP-BiomedBERT-large-uncased-abstract'\n",
    "model_short_name = 'BiomedBERT-large-abstract'\n",
    "run_number = '1'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.choice([5e-5, 2e-5, 1e-5, 8e-6, 5e-6, 2e-6, 1e-6]),\n",
    "    'num_train_epochs':tune.choice([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':8\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=40, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioBERT base-1.2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = 'dmis-lab/biobert-base-cased-v1.2'\n",
    "model_short_name = 'BioBERT-base'\n",
    "run_number = '1'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.choice([5e-5, 2e-5, 1e-5, 8e-6, 5e-6, 2e-6, 1e-6]),\n",
    "    'num_train_epochs':tune.choice([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':8\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=40, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioBERT large-1.1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = 'dmis-lab/biobert-large-cased-v1.1'\n",
    "model_short_name = 'BioBERT-large'\n",
    "run_number = '1'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.choice([5e-5, 2e-5, 1e-5, 8e-6, 5e-6, 2e-6, 1e-6]),\n",
    "    'num_train_epochs':tune.choice([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':8\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=40, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioBART v2-base"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = 'GanjinZero/biobart-v2-base'\n",
    "model_short_name = 'BioBART-base'\n",
    "run_number = '1'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.choice([5e-5, 2e-5, 1e-5, 8e-6, 5e-6, 2e-6, 1e-6]),\n",
    "    'num_train_epochs':tune.choice([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':8\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=40, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Test debug biobart"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BartForSequenceClassification were not initialized from the model checkpoint at GanjinZero/biobart-v2-base and are newly initialized: ['classification_head.dense.bias', 'classification_head.dense.weight', 'classification_head.out_proj.bias', 'classification_head.out_proj.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "/usr/local/lib/python3.11/site-packages/transformers/training_args.py:1545: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead\n",
      "  warnings.warn(\n",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='1204' max='1204' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [1204/1204 38:36, Epoch 4/4]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Step</th>\n",
       "      <th>Training Loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/transformers/configuration_utils.py:384: UserWarning: Some non-default generation parameters are set in the model config. These should go into either a) `model.generation_config` (as opposed to `model.config`); OR b) a GenerationConfig file (https://huggingface.co/docs/transformers/generation_strategies#save-a-custom-decoding-strategy-with-your-model).This warning will become an exception in the future.\n",
      "Non-default generation parameters: {'early_stopping': True, 'num_beams': 4, 'no_repeat_ngram_size': 3}\n",
      "  warnings.warn(\n",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n",
      "/usr/local/lib/python3.11/site-packages/transformers/configuration_utils.py:384: UserWarning: Some non-default generation parameters are set in the model config. These should go into either a) `model.generation_config` (as opposed to `model.config`); OR b) a GenerationConfig file (https://huggingface.co/docs/transformers/generation_strategies#save-a-custom-decoding-strategy-with-your-model).This warning will become an exception in the future.\n",
      "Non-default generation parameters: {'early_stopping': True, 'num_beams': 4, 'no_repeat_ngram_size': 3}\n",
      "  warnings.warn(\n",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n",
      "/usr/local/lib/python3.11/site-packages/transformers/configuration_utils.py:384: UserWarning: Some non-default generation parameters are set in the model config. These should go into either a) `model.generation_config` (as opposed to `model.config`); OR b) a GenerationConfig file (https://huggingface.co/docs/transformers/generation_strategies#save-a-custom-decoding-strategy-with-your-model).This warning will become an exception in the future.\n",
      "Non-default generation parameters: {'early_stopping': True, 'num_beams': 4, 'no_repeat_ngram_size': 3}\n",
      "  warnings.warn(\n",
      "/usr/local/lib/python3.11/site-packages/transformers/configuration_utils.py:384: UserWarning: Some non-default generation parameters are set in the model config. These should go into either a) `model.generation_config` (as opposed to `model.config`); OR b) a GenerationConfig file (https://huggingface.co/docs/transformers/generation_strategies#save-a-custom-decoding-strategy-with-your-model).This warning will become an exception in the future.\n",
      "Non-default generation parameters: {'early_stopping': True, 'num_beams': 4, 'no_repeat_ngram_size': 3}\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "model = AutoModelForSequenceClassification.from_pretrained('GanjinZero/biobart-v2-base', num_labels=2)\n",
    "training_args = TrainingArguments(filepath + '/models/test_debug/biobart-base', evaluation_strategy='no', logging_strategy='no', learning_rate=1e-5, num_train_epochs=4)\n",
    "trainer = Trainer(model=model, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train_val'])\n",
    "trainer.train()\n",
    "trainer.save_model(filepath + '/models/test_debug/biobart-base/model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='75' max='75' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [75/75 00:53]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "ename": "ValueError",
     "evalue": "setting an array element with a sequence. The requested array has an inhomogeneous shape after 2 dimensions. The detected shape was (2, 2400) + inhomogeneous part.",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[12], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[43mtrainer\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mevaluate\u001b[49m\u001b[43m(\u001b[49m\u001b[43meval_dataset\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mtokenized_datasets\u001b[49m\u001b[43m[\u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mtest\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m/usr/local/lib/python3.11/site-packages/transformers/trainer.py:3868\u001b[0m, in \u001b[0;36mTrainer.evaluate\u001b[0;34m(self, eval_dataset, ignore_keys, metric_key_prefix)\u001b[0m\n\u001b[1;32m   3865\u001b[0m start_time \u001b[38;5;241m=\u001b[39m time\u001b[38;5;241m.\u001b[39mtime()\n\u001b[1;32m   3867\u001b[0m eval_loop \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mprediction_loop \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39margs\u001b[38;5;241m.\u001b[39muse_legacy_prediction_loop \u001b[38;5;28;01melse\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mevaluation_loop\n\u001b[0;32m-> 3868\u001b[0m output \u001b[38;5;241m=\u001b[39m \u001b[43meval_loop\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m   3869\u001b[0m \u001b[43m    \u001b[49m\u001b[43meval_dataloader\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   3870\u001b[0m \u001b[43m    \u001b[49m\u001b[43mdescription\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[38;5;124;43mEvaluation\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\n\u001b[1;32m   3871\u001b[0m \u001b[43m    \u001b[49m\u001b[38;5;66;43;03m# No point gathering the predictions if there are no metrics, otherwise we defer to\u001b[39;49;00m\n\u001b[1;32m   3872\u001b[0m \u001b[43m    \u001b[49m\u001b[38;5;66;43;03m# self.args.prediction_loss_only\u001b[39;49;00m\n\u001b[1;32m   3873\u001b[0m \u001b[43m    \u001b[49m\u001b[43mprediction_loss_only\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43;01mTrue\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mif\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mcompute_metrics\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;129;43;01mis\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mNone\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[38;5;28;43;01melse\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mNone\u001b[39;49;00m\u001b[43m,\u001b[49m\n\u001b[1;32m   3874\u001b[0m \u001b[43m    \u001b[49m\u001b[43mignore_keys\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mignore_keys\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   3875\u001b[0m \u001b[43m    \u001b[49m\u001b[43mmetric_key_prefix\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mmetric_key_prefix\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   3876\u001b[0m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   3878\u001b[0m total_batch_size \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39margs\u001b[38;5;241m.\u001b[39meval_batch_size \u001b[38;5;241m*\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39margs\u001b[38;5;241m.\u001b[39mworld_size\n\u001b[1;32m   3879\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mmetric_key_prefix\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m_jit_compilation_time\u001b[39m\u001b[38;5;124m\"\u001b[39m \u001b[38;5;129;01min\u001b[39;00m output\u001b[38;5;241m.\u001b[39mmetrics:\n",
      "File \u001b[0;32m/usr/local/lib/python3.11/site-packages/transformers/trainer.py:4160\u001b[0m, in \u001b[0;36mTrainer.evaluation_loop\u001b[0;34m(self, dataloader, description, prediction_loss_only, ignore_keys, metric_key_prefix)\u001b[0m\n\u001b[1;32m   4156\u001b[0m         metrics \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mcompute_metrics(\n\u001b[1;32m   4157\u001b[0m             EvalPrediction(predictions\u001b[38;5;241m=\u001b[39mall_preds, label_ids\u001b[38;5;241m=\u001b[39mall_labels, inputs\u001b[38;5;241m=\u001b[39mall_inputs)\n\u001b[1;32m   4158\u001b[0m         )\n\u001b[1;32m   4159\u001b[0m     \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m-> 4160\u001b[0m         metrics \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mcompute_metrics\u001b[49m\u001b[43m(\u001b[49m\u001b[43mEvalPrediction\u001b[49m\u001b[43m(\u001b[49m\u001b[43mpredictions\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mall_preds\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mlabel_ids\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mall_labels\u001b[49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   4161\u001b[0m \u001b[38;5;28;01melif\u001b[39;00m metrics \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m   4162\u001b[0m     metrics \u001b[38;5;241m=\u001b[39m {}\n",
      "Cell \u001b[0;32mIn[9], line 3\u001b[0m, in \u001b[0;36mcompute_single_label_metrics\u001b[0;34m(eval_pred)\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mcompute_single_label_metrics\u001b[39m(eval_pred):\n\u001b[1;32m      2\u001b[0m     logits, labels \u001b[38;5;241m=\u001b[39m eval_pred\n\u001b[0;32m----> 3\u001b[0m     predictions \u001b[38;5;241m=\u001b[39m \u001b[43mnp\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43margmax\u001b[49m\u001b[43m(\u001b[49m\u001b[43mlogits\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43maxis\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;241;43m1\u001b[39;49m\u001b[43m)\u001b[49m\n\u001b[1;32m      4\u001b[0m     metrics \u001b[38;5;241m=\u001b[39m {\u001b[38;5;124m'\u001b[39m\u001b[38;5;124maccuracy\u001b[39m\u001b[38;5;124m'\u001b[39m:accuracy_score(labels, predictions), \u001b[38;5;124m'\u001b[39m\u001b[38;5;124mf1\u001b[39m\u001b[38;5;124m'\u001b[39m:f1_score(labels, predictions), \u001b[38;5;124m'\u001b[39m\u001b[38;5;124mprecision\u001b[39m\u001b[38;5;124m'\u001b[39m:precision_score(labels, predictions), \u001b[38;5;124m'\u001b[39m\u001b[38;5;124mrecall\u001b[39m\u001b[38;5;124m'\u001b[39m:recall_score(labels, predictions)}\n\u001b[1;32m      5\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m metrics\n",
      "File \u001b[0;32m/usr/local/lib/python3.11/site-packages/numpy/core/fromnumeric.py:1229\u001b[0m, in \u001b[0;36margmax\u001b[0;34m(a, axis, out, keepdims)\u001b[0m\n\u001b[1;32m   1142\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m   1143\u001b[0m \u001b[38;5;124;03mReturns the indices of the maximum values along an axis.\u001b[39;00m\n\u001b[1;32m   1144\u001b[0m \n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   1226\u001b[0m \u001b[38;5;124;03m(2, 1, 4)\u001b[39;00m\n\u001b[1;32m   1227\u001b[0m \u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m   1228\u001b[0m kwds \u001b[38;5;241m=\u001b[39m {\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mkeepdims\u001b[39m\u001b[38;5;124m'\u001b[39m: keepdims} \u001b[38;5;28;01mif\u001b[39;00m keepdims \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m np\u001b[38;5;241m.\u001b[39m_NoValue \u001b[38;5;28;01melse\u001b[39;00m {}\n\u001b[0;32m-> 1229\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43m_wrapfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[43ma\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43margmax\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43maxis\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43maxis\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mout\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mout\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwds\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m/usr/local/lib/python3.11/site-packages/numpy/core/fromnumeric.py:56\u001b[0m, in \u001b[0;36m_wrapfunc\u001b[0;34m(obj, method, *args, **kwds)\u001b[0m\n\u001b[1;32m     54\u001b[0m bound \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mgetattr\u001b[39m(obj, method, \u001b[38;5;28;01mNone\u001b[39;00m)\n\u001b[1;32m     55\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m bound \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[0;32m---> 56\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43m_wrapit\u001b[49m\u001b[43m(\u001b[49m\u001b[43mobj\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mmethod\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwds\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     58\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m     59\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m bound(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwds)\n",
      "File \u001b[0;32m/usr/local/lib/python3.11/site-packages/numpy/core/fromnumeric.py:45\u001b[0m, in \u001b[0;36m_wrapit\u001b[0;34m(obj, method, *args, **kwds)\u001b[0m\n\u001b[1;32m     43\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mAttributeError\u001b[39;00m:\n\u001b[1;32m     44\u001b[0m     wrap \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[0;32m---> 45\u001b[0m result \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mgetattr\u001b[39m(asarray(obj), method)(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwds)\n\u001b[1;32m     46\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m wrap:\n\u001b[1;32m     47\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(result, mu\u001b[38;5;241m.\u001b[39mndarray):\n",
      "\u001b[0;31mValueError\u001b[0m: setting an array element with a sequence. The requested array has an inhomogeneous shape after 2 dimensions. The detected shape was (2, 2400) + inhomogeneous part."
     ]
    }
   ],
   "source": [
    "trainer.evaluate(eval_dataset=tokenized_datasets['test'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioBART v2-large"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioELECTRA base-discriminator-pubmed"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### First test that it will work \n",
    "(avoid issue similar to BioBART)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "262782af72a2494db22423b4fcf171d8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "config.json:   0%|          | 0.00/665 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "ename": "OSError",
     "evalue": "kamalkraj/bioelectra-base-discriminator-pubmed does not appear to have a file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt or flax_model.msgpack.",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mOSError\u001b[0m                                   Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[13], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m model \u001b[38;5;241m=\u001b[39m \u001b[43mAutoModelForSequenceClassification\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mfrom_pretrained\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mkamalkraj/bioelectra-base-discriminator-pubmed\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mnum_labels\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;241;43m2\u001b[39;49m\u001b[43m)\u001b[49m\n\u001b[1;32m      2\u001b[0m training_args \u001b[38;5;241m=\u001b[39m TrainingArguments(filepath \u001b[38;5;241m+\u001b[39m \u001b[38;5;124m'\u001b[39m\u001b[38;5;124m/models/test_debug/bioelectra-base\u001b[39m\u001b[38;5;124m'\u001b[39m, evaluation_strategy\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mno\u001b[39m\u001b[38;5;124m'\u001b[39m, logging_strategy\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mno\u001b[39m\u001b[38;5;124m'\u001b[39m, learning_rate\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m1e-5\u001b[39m, num_train_epochs\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m4\u001b[39m)\n\u001b[1;32m      3\u001b[0m trainer \u001b[38;5;241m=\u001b[39m Trainer(model\u001b[38;5;241m=\u001b[39mmodel, args\u001b[38;5;241m=\u001b[39mtraining_args, compute_metrics\u001b[38;5;241m=\u001b[39mcompute_single_label_metrics, train_dataset\u001b[38;5;241m=\u001b[39mtokenized_datasets[\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mtrain_val\u001b[39m\u001b[38;5;124m'\u001b[39m])\n",
      "File \u001b[0;32m/usr/local/lib/python3.11/site-packages/transformers/models/auto/auto_factory.py:564\u001b[0m, in \u001b[0;36m_BaseAutoModelClass.from_pretrained\u001b[0;34m(cls, pretrained_model_name_or_path, *model_args, **kwargs)\u001b[0m\n\u001b[1;32m    562\u001b[0m \u001b[38;5;28;01melif\u001b[39;00m \u001b[38;5;28mtype\u001b[39m(config) \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28mcls\u001b[39m\u001b[38;5;241m.\u001b[39m_model_mapping\u001b[38;5;241m.\u001b[39mkeys():\n\u001b[1;32m    563\u001b[0m     model_class \u001b[38;5;241m=\u001b[39m _get_model_class(config, \u001b[38;5;28mcls\u001b[39m\u001b[38;5;241m.\u001b[39m_model_mapping)\n\u001b[0;32m--> 564\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mmodel_class\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mfrom_pretrained\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    565\u001b[0m \u001b[43m        \u001b[49m\u001b[43mpretrained_model_name_or_path\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mmodel_args\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mconfig\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mconfig\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mhub_kwargs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\n\u001b[1;32m    566\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    567\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(\n\u001b[1;32m    568\u001b[0m     \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mUnrecognized configuration class \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mconfig\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__class__\u001b[39m\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m for this kind of AutoModel: \u001b[39m\u001b[38;5;132;01m{\u001b[39;00m\u001b[38;5;28mcls\u001b[39m\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__name__\u001b[39m\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m.\u001b[39m\u001b[38;5;130;01m\\n\u001b[39;00m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    569\u001b[0m     \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mModel type should be one of \u001b[39m\u001b[38;5;132;01m{\u001b[39;00m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m, \u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;241m.\u001b[39mjoin(c\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__name__\u001b[39m\u001b[38;5;250m \u001b[39m\u001b[38;5;28;01mfor\u001b[39;00m\u001b[38;5;250m \u001b[39mc\u001b[38;5;250m \u001b[39m\u001b[38;5;129;01min\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[38;5;28mcls\u001b[39m\u001b[38;5;241m.\u001b[39m_model_mapping\u001b[38;5;241m.\u001b[39mkeys())\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m.\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    570\u001b[0m )\n",
      "File \u001b[0;32m/usr/local/lib/python3.11/site-packages/transformers/modeling_utils.py:3708\u001b[0m, in \u001b[0;36mPreTrainedModel.from_pretrained\u001b[0;34m(cls, pretrained_model_name_or_path, config, cache_dir, ignore_mismatched_sizes, force_download, local_files_only, token, revision, use_safetensors, *model_args, **kwargs)\u001b[0m\n\u001b[1;32m   3702\u001b[0m                 \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mEnvironmentError\u001b[39;00m(\n\u001b[1;32m   3703\u001b[0m                     \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mpretrained_model_name_or_path\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m does not appear to have a file named\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m   3704\u001b[0m                     \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m \u001b[39m\u001b[38;5;132;01m{\u001b[39;00m_add_variant(WEIGHTS_NAME,\u001b[38;5;250m \u001b[39mvariant)\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m but there is a file without the variant\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m   3705\u001b[0m                     \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mvariant\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m. Use `variant=None` to load this model from those weights.\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m   3706\u001b[0m                 )\n\u001b[1;32m   3707\u001b[0m             \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m-> 3708\u001b[0m                 \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mEnvironmentError\u001b[39;00m(\n\u001b[1;32m   3709\u001b[0m                     \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mpretrained_model_name_or_path\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m does not appear to have a file named\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m   3710\u001b[0m                     \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m \u001b[39m\u001b[38;5;132;01m{\u001b[39;00m_add_variant(WEIGHTS_NAME,\u001b[38;5;250m \u001b[39mvariant)\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m, \u001b[39m\u001b[38;5;132;01m{\u001b[39;00m_add_variant(SAFE_WEIGHTS_NAME,\u001b[38;5;250m \u001b[39mvariant)\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m,\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m   3711\u001b[0m                     \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mTF2_WEIGHTS_NAME\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m, \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mTF_WEIGHTS_NAME\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m or \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mFLAX_WEIGHTS_NAME\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m.\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m   3712\u001b[0m                 )\n\u001b[1;32m   3714\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mEnvironmentError\u001b[39;00m:\n\u001b[1;32m   3715\u001b[0m     \u001b[38;5;66;03m# Raise any environment error raise by `cached_file`. It will have a helpful error message adapted\u001b[39;00m\n\u001b[1;32m   3716\u001b[0m     \u001b[38;5;66;03m# to the original exception.\u001b[39;00m\n\u001b[1;32m   3717\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m\n",
      "\u001b[0;31mOSError\u001b[0m: kamalkraj/bioelectra-base-discriminator-pubmed does not appear to have a file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt or flax_model.msgpack."
     ]
    }
   ],
   "source": [
    "model = AutoModelForSequenceClassification.from_pretrained('kamalkraj/bioelectra-base-discriminator-pubmed', num_labels=2)\n",
    "training_args = TrainingArguments(filepath + '/models/test_debug/bioelectra-base', evaluation_strategy='no', logging_strategy='no', learning_rate=1e-5, num_train_epochs=4)\n",
    "trainer = Trainer(model=model, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train_val'])\n",
    "trainer.train()\n",
    "trainer.save_model(filepath + '/models/test_debug/bioelectra-base/model')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioELECTRA base-discriminator-pubmed-pmc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioELECTRA base-discriminator-pubmed-pmc-lt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioGPT\n",
    "\n",
    "Setting batch_size to 4 to see if it works, with 8 all trials end in out of memory error."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = 'microsoft/biogpt'\n",
    "model_short_name = 'BioGPT'\n",
    "run_number = '1'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.choice([5e-5, 2e-5, 1e-5, 8e-6, 5e-6, 2e-6, 1e-6]),\n",
    "    'num_train_epochs':tune.choice([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':4\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=40, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioGPT-large"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BioGPT-large-PubMedQA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run2- Further explore the hyperparameter space of the models who rated best\n",
    "\n",
    "- BiomedBERT-base-abstract 8e-6x4\n",
    "- BiomedBERT-base-abstract 1e-5x2\n",
    "- BiomedBERT-base-abstract 2e-5x4\n",
    "- BiomedBERT-base-abstract 2e-5x4\n",
    "\n",
    "\n",
    "- BiomedBERT-large-abstract 8e-6x2\n",
    "- BiomedBERT-large-abstract 8e-6x4\n",
    "- BiomedBERT-large-abstract 2e-5x4\n",
    "\n",
    "\n",
    "Results: BiomedBERT-base abstract 2e-5x3 seems to give the most consistent results with recall of 0.86 and precision of 0.78-0.79-0.8"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BiomedBERT-base-abstract\n",
    "\n",
    "BiomedBERT base-abstract // microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = 'microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract'\n",
    "model_short_name = 'BiomedBERT-base-abstract'\n",
    "run_number = '2'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.grid_search([5e-5, 2e-5, 1e-5, 8e-6, 5e-6]),\n",
    "    'num_train_epochs':tune.grid_search([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':8\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=3, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### BiomedBERT-large-abstract\n",
    "\n",
    "BiomedBERT large-abstract // microsoft/BiomedNLP-BiomedBERT-large-uncased-abstract"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e642a0c6a6b54b5a94161a37d40f6b98",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "tokenizer_config.json:   0%|          | 0.00/27.0 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "813958cf1c49418585e8095e261d0416",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "config.json:   0%|          | 0.00/1.01k [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "872961d032fe4523820793b1057926f1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "vocab.txt:   0%|          | 0.00/225k [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "918f46eb71574df6bf2c4bf6e5637baa",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/9624 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "99a5c28b25d448c9b39423c952c2e2e1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/7699 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6d78c269df834dd5910b0ff1ea016b70",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/1925 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1e6b57f80fca444f94a08c6d4402d94c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/2400 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4c602cbb1486448191e06d0ad2375863",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "pytorch_model.bin:   0%|          | 0.00/1.34G [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = 'microsoft/BiomedNLP-BiomedBERT-large-uncased-abstract'\n",
    "model_short_name = 'BiomedBERT-large-abstract'\n",
    "run_number = '2'\n",
    "saving_dir_name = 'run' + run_number + '/' + model_short_name + '-run' + run_number\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model)\n",
    "\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "def model_init():\n",
    "    return AutoModelForSequenceClassification.from_pretrained(model, num_labels=2)\n",
    "\n",
    "hp_space = {\n",
    "    'learning_rate':tune.grid_search([5e-5, 2e-5, 1e-5, 8e-6, 5e-6]),\n",
    "    'num_train_epochs':tune.grid_search([1,2,3,4,5,6]),\n",
    "    'per_device_train_batch_size':8\n",
    "}\n",
    "\n",
    "training_args = TrainingArguments(filepath + '/results/train_results/' + saving_dir_name, eval_strategy='epoch', save_strategy='no', disable_tqdm=True)\n",
    "trainer = Trainer(model_init=model_init, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train'], eval_dataset=tokenized_datasets['val'])\n",
    "best_model = trainer.hyperparameter_search(hp_space=lambda _: hp_space, direction='maximize', n_trials=3, storage_path=filepath + '/results/tune_results/', name=saving_dir_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluate the selected model on the test set without threshold adjustment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract 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"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fdcccf5a6e2949d099c2cf8d20e0dfe1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/9624 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3854ba591dda4d7288d3f68898162d9c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/7699 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9f981fefbc794c7d98b71c8912e9a041",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/1925 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a0b4b3460dd048a3b0debe33203b0312",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/2400 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/transformers/training_args.py:1545: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead\n",
      "  warnings.warn(\n",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='903' max='903' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [903/903 22:59, Epoch 3/3]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Step</th>\n",
       "      <th>Training Loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    }
   ],
   "source": [
    "model_name = 'microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract'\n",
    "\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
    "model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2)\n",
    "tokenized_datasets = ds.map(tokenizer_function, batched=True)\n",
    "training_args = TrainingArguments(filepath + '/models/test_set_eval_models/inclusion_biomedbert', eval_strategy='no', logging_strategy='no', learning_rate=2e-5, num_train_epochs=3)\n",
    "trainer = Trainer(model=model, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_datasets['train_val'])\n",
    "trainer.train()\n",
    "trainer.save_model(filepath + '/models/test_set_eval_models/inclusion_biomedbert/model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='75' max='75' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [75/75 00:39]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'eval_loss': 0.08033862709999084,\n",
       " 'eval_accuracy': 0.9758333333333333,\n",
       " 'eval_f1': 0.7363636363636363,\n",
       " 'eval_precision': 0.680672268907563,\n",
       " 'eval_recall': 0.801980198019802,\n",
       " 'eval_runtime': 39.8548,\n",
       " 'eval_samples_per_second': 60.219,\n",
       " 'eval_steps_per_second': 1.882,\n",
       " 'epoch': 3.0}"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trainer.evaluate(eval_dataset=tokenized_datasets['test'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Determine thresholds to automatically exclude papers\n",
    "Goal: Recall = 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = 'microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract'\n",
    "\n",
    "all_df = df[df_columns_to_keep]\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
    "tokenizer_kwargs = {'padding':True,'truncation':True,'max_length':512}\n",
    "\n",
    "def tokenizer_function(samples):\n",
    "    return tokenizer(samples['text'], padding=\"max_length\", truncation=True, max_length=512)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                text  labels\n",
      "2  Leveraging the pharmacy executive beyond the p...     0.0\n",
      "3  The utilization of simulated patients for teac...     0.0\n",
      "5  Exploration of pharmacist comfort with harm re...     0.0\n",
      "6  Challenges for Polish Community Pharmacists in...     0.0\n",
      "7  Pharmacists' Perceptions of Handling Antimicro...     0.0\n",
      "                                                 text  labels\n",
      "1   The utility of Conversation Analysis versus Ro...     0.0\n",
      "4   Implementing an Antimicrobial Stewardship Prog...     0.0\n",
      "9   Development and Implementation of In-House Pha...     0.0\n",
      "11  A multifaceted approach in leveraging electron...     0.0\n",
      "15  Identifying COVID-19 and H1N1 vaccination hesi...     0.0\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5a90691aeec84f4094a2351b1087263a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6a87b679f62b4f2e97160447285f71d9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "140c6cf0a9234474860ac56f532ec07c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract 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",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='903' max='903' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [903/903 22:58, Epoch 3/3]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Step</th>\n",
       "      <th>Training Loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/transformers/pipelines/text_classification.py:104: UserWarning: `return_all_scores` is now deprecated,  if want a similar functionality use `top_k=None` instead of `return_all_scores=True` or `top_k=1` instead of `return_all_scores=False`.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0b649e2071d747c9ba3fe5fd189c9549",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2405 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "You seem to be using the pipelines sequentially on GPU. In order to maximize efficiency please use a dataset\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                text  labels\n",
      "1  The utility of Conversation Analysis versus Ro...     0.0\n",
      "2  Leveraging the pharmacy executive beyond the p...     0.0\n",
      "3  The utilization of simulated patients for teac...     0.0\n",
      "4  Implementing an Antimicrobial Stewardship Prog...     0.0\n",
      "5  Exploration of pharmacist comfort with harm re...     0.0\n",
      "                                                 text  labels\n",
      "13  Implementing a personalized pharmaceutical pla...     0.0\n",
      "21  Development and Implementation of Interprofess...     0.0\n",
      "24  Prescribing of Pressor Agents in Septic Shock:...     0.0\n",
      "31  Strategy to Improve Efficiency of Comprehensiv...     0.0\n",
      "33  Impact of a Clinical Decision-Support Tool on ...     0.0\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1cbe8592d87c4c89b0a20c80cce804dd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7712132a0d794dd0beef4fb30fd96dc6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d219d848eb9646a9b939e6cd587fa3d0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract 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",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='903' max='903' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [903/903 22:57, Epoch 3/3]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Step</th>\n",
       "      <th>Training Loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/transformers/pipelines/text_classification.py:104: UserWarning: `return_all_scores` is now deprecated,  if want a similar functionality use `top_k=None` instead of `return_all_scores=True` or `top_k=1` instead of `return_all_scores=False`.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dadcc1912fc74ca4af0eb133c537323c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2405 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                text  labels\n",
      "1  The utility of Conversation Analysis versus Ro...     0.0\n",
      "2  Leveraging the pharmacy executive beyond the p...     0.0\n",
      "3  The utilization of simulated patients for teac...     0.0\n",
      "4  Implementing an Antimicrobial Stewardship Prog...     0.0\n",
      "5  Exploration of pharmacist comfort with harm re...     0.0\n",
      "                                                 text  labels\n",
      "16  Economic impact and health care utilization ou...     0.0\n",
      "27  An Evaluation of Pharmacist-Led Interventions ...     1.0\n",
      "28  The Unmet Need for Pharmacists as Important Me...     0.0\n",
      "29  Team Communication on Tasks Performed in Blood...     0.0\n",
      "35  Preventing Prescription Drug Abuse: A Pharmacy...     0.0\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5360235f3d8d4b45be9c8c8b07336ea7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "370354c92a51488ca915c61911cb6223",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "631b904be22d46df8a090585338027aa",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract 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",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='903' max='903' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [903/903 22:55, Epoch 3/3]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Step</th>\n",
       "      <th>Training Loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/transformers/pipelines/text_classification.py:104: UserWarning: `return_all_scores` is now deprecated,  if want a similar functionality use `top_k=None` instead of `return_all_scores=True` or `top_k=1` instead of `return_all_scores=False`.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4cded673a7ee4eb9b9586b51a84ee6d2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2405 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                text  labels\n",
      "1  The utility of Conversation Analysis versus Ro...     0.0\n",
      "2  Leveraging the pharmacy executive beyond the p...     0.0\n",
      "4  Implementing an Antimicrobial Stewardship Prog...     0.0\n",
      "5  Exploration of pharmacist comfort with harm re...     0.0\n",
      "6  Challenges for Polish Community Pharmacists in...     0.0\n",
      "                                                 text  labels\n",
      "3   The utilization of simulated patients for teac...     0.0\n",
      "7   Pharmacists' Perceptions of Handling Antimicro...     0.0\n",
      "8   Metabolic Activation of the Toxic Natural Prod...     0.0\n",
      "19  The psychological health of patients diagnosed...     0.0\n",
      "23  Comparison of Interventions Made in an Ambulat...     1.0\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1ac2eeb7c19243aeaa270b56160fccd5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6f0c0edfe1a64405b9be3c8549317521",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b2ecd42f8cd147988ad58cf3d28e9729",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/9619 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract 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",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='903' max='903' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [903/903 22:56, Epoch 3/3]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Step</th>\n",
       "      <th>Training Loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/transformers/pipelines/text_classification.py:104: UserWarning: `return_all_scores` is now deprecated,  if want a similar functionality use `top_k=None` instead of `return_all_scores=True` or `top_k=1` instead of `return_all_scores=False`.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a5bdb812f44548599bf487ad56fa42b2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2405 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                                text  labels\n",
      "1  The utility of Conversation Analysis versus Ro...     0.0\n",
      "3  The utilization of simulated patients for teac...     0.0\n",
      "4  Implementing an Antimicrobial Stewardship Prog...     0.0\n",
      "7  Pharmacists' Perceptions of Handling Antimicro...     0.0\n",
      "8  Metabolic Activation of the Toxic Natural Prod...     0.0\n",
      "                                                 text  labels\n",
      "2   Leveraging the pharmacy executive beyond the p...     0.0\n",
      "5   Exploration of pharmacist comfort with harm re...     0.0\n",
      "6   Challenges for Polish Community Pharmacists in...     0.0\n",
      "10  Pharmacists' perceived role in supporting diab...     0.0\n",
      "12  Decision coaching for people making healthcare...     0.0\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "83df82a37ce74e8fa6ddf36b249192f5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Stringifying the column:   0%|          | 0/9620 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bfd7219b2f0549a5b0ef9bb97f9b061d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Casting to class labels:   0%|          | 0/9620 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "aa2449bce0fb413cbc2b297a7789266d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/9620 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract 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",
      "/usr/local/lib/python3.11/site-packages/torch/nn/parallel/_functions.py:68: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n",
      "  warnings.warn('Was asked to gather along dimension 0, but all '\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='903' max='903' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [903/903 22:55, Epoch 3/3]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Step</th>\n",
       "      <th>Training Loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.11/site-packages/transformers/pipelines/text_classification.py:104: UserWarning: `return_all_scores` is now deprecated,  if want a similar functionality use `top_k=None` instead of `return_all_scores=True` or `top_k=1` instead of `return_all_scores=False`.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a2042af9abf64fc7a0c4da14f291a435",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2404 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "kf = KFold(n_splits=5, shuffle=True)\n",
    "thresholds = []\n",
    "\n",
    "for train_idx, val_idx in kf.split(all_df):\n",
    "    train = all_df.iloc[train_idx]\n",
    "    print(train.head())\n",
    "    val = all_df.iloc[val_idx]\n",
    "    print(val.head())\n",
    "    ds = Dataset.from_pandas(train)\n",
    "    ds.cleanup_cache_files()\n",
    "    ds = ds.class_encode_column('labels')\n",
    "\n",
    "    tokenized_ds = ds.map(tokenizer_function, batched=True)\n",
    "\n",
    "    model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2)\n",
    "    training_args = TrainingArguments(filepath + '/.temp', save_strategy='no', eval_strategy='no', logging_strategy='no', overwrite_output_dir=True, learning_rate=2e-5, num_train_epochs=3)\n",
    "    trainer = Trainer(model=model, args=training_args, compute_metrics=compute_single_label_metrics, train_dataset=tokenized_ds)\n",
    "    trainer.train()\n",
    "\n",
    "    pipe = TextClassificationPipeline(model=model, tokenizer=tokenizer, top_k=None, device=0)\n",
    "    scores = []\n",
    "    for i in trange(len(val)):\n",
    "        scores.append(pipe(val.iloc[i]['text'], **tokenizer_kwargs)[0][1]['score'])\n",
    "\n",
    "    PrecisionRecallDisplay.from_predictions(val['labels'], scores)\n",
    "\n",
    "    pr,re,th = precision_recall_curve(val['labels'], scores)\n",
    "    rth_df = pd.DataFrame.from_dict({'precision':pr[:-1],'recall':re[:-1],'threshold':th})\n",
    "\n",
    "    ceiling_recall = 0.99\n",
    "    thresholds.append(rth_df[rth_df.recall >= ceiling_recall].iloc[-1]['threshold'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000509</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.000510</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000687</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.000518</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.001695</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          0\n",
       "0  0.000509\n",
       "1  0.000510\n",
       "2  0.000687\n",
       "3  0.000518\n",
       "4  0.001695"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "thr_df = pd.DataFrame(thresholds)\n",
    "thr_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean threshold: 0.00078370631672441959\n",
      "Number of abstracts excluded from review based on recall threshold: 1908 out of 2404 for ratio of 0.794\n",
      "Adjusted threshold precision: 0.242, recall: 1.000, f1: 0.390, accuracy: 0.844\n",
      "Number of incorrectly excluded abstracts: 0\n"
     ]
    }
   ],
   "source": [
    "threshold = thr_df.mean()[0]\n",
    "print('Mean threshold: {:.20f}'.format(threshold))\n",
    "recall_test = []\n",
    "preds = []\n",
    "incorrect = []\n",
    "for i, label, score in zip(range(len(val['labels'])), val['labels'], scores):\n",
    "    if score >= threshold:\n",
    "        recall_test.append({'index':i, 'label':label, 'score':score})\n",
    "        preds.append(1)\n",
    "    else:\n",
    "        if label == 1:\n",
    "            incorrect.append(i)\n",
    "        preds.append(0)\n",
    "print('Number of abstracts excluded from review based on recall threshold: {} out of {} for ratio of {:.3f}'.format(len(val) - len(recall_test), len(val), (len(val) - len(recall_test)) / len(val)))\n",
    "print('Adjusted threshold precision: {:.3f}, recall: {:.3f}, f1: {:.3f}, accuracy: {:.3f}'.format(precision_score(val['labels'], preds), recall_score(val['labels'], preds), f1_score(val['labels'], preds), accuracy_score(val['labels'], preds)))\n",
    "print('Number of incorrectly excluded abstracts: {}'.format(len(incorrect)))\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.11.4"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}