test.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "900c37c1",
   "metadata": {},
   "source": [
    "**For Test**\n",
    "===\n",
    "> -*- coding: utf-8 -*-\n",
    "\n",
    "> Author: xinghd"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "294d137c",
   "metadata": {},
   "source": [
    "## Import libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "2b0a31dc",
   "metadata": {},
   "outputs": [],
   "source": [
    "from utils.utils import Trainer, Predictor, TrainReader,TestReader,PredictReader, data_collate, get_shuffle_data, collate_fn, seed_everything\n",
    "from model.model import Encoder, Decoder, ModelCat\n",
    "\n",
    "%reload_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "2860079c",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import json\n",
    "import random\n",
    "import time\n",
    "import timeit\n",
    "import pickle\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from rdkit import Chem\n",
    "from rdkit.Chem import Descriptors\n",
    "from torch import nn\n",
    "from sklearn.model_selection import train_test_split, StratifiedKFold, KFold\n",
    "import torch\n",
    "from torch.utils.data import DataLoader, random_split\n",
    "from utils.word2vec import seq_to_kmers, get_protein_embedding\n",
    "from gensim.models import Word2Vec\n",
    "from torch.utils import data as torch_data\n",
    "import warnings"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cda0b765",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "6db220de",
   "metadata": {},
   "outputs": [],
   "source": [
    "assert torch.cuda.is_available(), \"Must have a avaliable gpu\"\n",
    "\n",
    "class CFG :\n",
    "    IFwandb = False\n",
    "    # Data root\n",
    "    DATA_ROOT = r'./data/'\n",
    "    # Save unlabel data that may be positive\n",
    "    U_m_P_savepath = r\"./data/U_m_P/\"\n",
    "    # Word2vec model path\n",
    "    word2vec_path = './model/model_pretrained/word2vec_pretrained.model' \n",
    "    # Prediction model path\n",
    "    state_dict_path ='./model/model_pretrained/PU-EPP_pretrained.pt'\n",
    "    # Number CUDA Devices:\n",
    "    gpu_number = torch.cuda.device_count()\n",
    "    # Device\n",
    "    DEVICE = torch.device('cuda:0')\n",
    "    # ====================================================\n",
    "    # Hyperparameters\n",
    "    # ====================================================\n",
    "    # Batch size\n",
    "    BATCH_SIZE = 3\n",
    "    # Epoch\n",
    "    EPOCHES = 100\n",
    "    # Block layers\n",
    "    layer_num = 12\n",
    "    # The last dimension of the protein data\n",
    "    protein_dim = 100\n",
    "    # The last dimension of the compound data\n",
    "    atom_dim = 46\n",
    "    # Hidden dimension\n",
    "    hid_dim = 128\n",
    "    # Norm_shape: layernorm parameter\n",
    "    norm_shape = 128\n",
    "    # ====================================================\n",
    "    # To create trainreader\n",
    "    # ====================================================\n",
    "    # If use pu learning\n",
    "    ifpu = True\n",
    "    # If use label smoothing\n",
    "    ifsmoothing = True\n",
    "    # ====================================================\n",
    "    # To create trainer\n",
    "    # ====================================================\n",
    "    # Learning rate\n",
    "    lr = 1e-4\n",
    "    # weight_decay\n",
    "    weight_decay = 2e-4\n",
    "    # Start removing potential postives from unlabeled samples when AUC > del_threshold\n",
    "    del_threshold=0.9\n",
    "    \n",
    "    # create result file\n",
    "    result_file_path = './results/log/'\n",
    "    # save_best_model(depends on AUC)\n",
    "    best_model_savepath = \"./model/model_save/\"\n",
    "    # \n",
    "    modelsave_file_suffix = 'example_epoch.pt'\n",
    "    result_file_suffix = 'example_log.txt'\n",
    "    \n",
    "    quantile = 0.9\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "feaa424f",
   "metadata": {},
   "source": [
    "## Parameters"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc5e64bf",
   "metadata": {},
   "source": [
    "### Random Seed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "3299feb0",
   "metadata": {},
   "outputs": [],
   "source": [
    "seed_everything(seed=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "84547467",
   "metadata": {},
   "source": [
    "## Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "4e8402f3",
   "metadata": {},
   "outputs": [],
   "source": [
    "encoder = Encoder(CFG.protein_dim, CFG.hid_dim, CFG.norm_shape)\n",
    "decoder = Decoder(CFG.atom_dim, CFG.hid_dim, CFG.norm_shape)\n",
    "model = ModelCat(encoder, decoder, atom_dim=CFG.atom_dim)\n",
    "model = model.to(CFG.DEVICE)\n",
    "model = nn.DataParallel(model, device_ids = list(range(CFG.gpu_number)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "a9abf947",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<All keys matched successfully>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.load_state_dict(torch.load(CFG.state_dict_path))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "943f0327",
   "metadata": {},
   "source": [
    "## Trainer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3a932341",
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, CFG)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fbd0f443",
   "metadata": {},
   "source": [
    "## Dataset & Dataloader"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c6674e61",
   "metadata": {},
   "source": [
    "### Dataset \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "574e32ad",
   "metadata": {},
   "outputs": [],
   "source": [
    "testdata = pd.read_csv('./data/PU-EPP_testdata.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cbff8bcd",
   "metadata": {},
   "source": [
    "### Dataloader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "ea834f6e",
   "metadata": {},
   "outputs": [],
   "source": [
    "testdata = TestReader(data=testdata, word2vec_path=CFG.word2vec_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "be5d34f3",
   "metadata": {},
   "outputs": [],
   "source": [
    "val_dataloader = DataLoader(testdata, batch_size=CFG.BATCH_SIZE,collate_fn=collate_fn)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e161fef1",
   "metadata": {},
   "source": [
    "## Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "23271d33-460e-40d3-b8a2-b2aa13acec76",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "AUC_dev, precision, recall, PRC_dev = trainer.test(val_dataloader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "f6f57580-3edc-451e-bccc-4aa69c6180bc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.9854439100000001, 0.9672081525577642, 0.9586, 0.9883653065367067)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "AUC_dev, precision, recall, PRC_dev"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "forpymol",
   "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": "3.7.13"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  },
  "vscode": {
   "interpreter": {
    "hash": "31f2aee4e71d21fbe5cf8b01ff0e069b9275f58929596ceb00d14d90e3e16cd6"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}