Deleting content

"""

Tests to make sure TF models can overfit on tiny datasets.

"""

from __future__ import print_function

from __future__ import division

from __future__ import unicode_literals

__author__ = "Bharath Ramsundar"

__copyright__ = "Copyright 2016, Stanford University"

__license__ = "LGPL"

import tempfile

import numpy as np

import unittest

from deepchem import metrics

from deepchem.datasets import Dataset

from deepchem.metrics import Metric

from deepchem.models.test import TestAPI

from deepchem.utils.evaluate import Evaluator

from deepchem.models.tensorflow_models import TensorflowModel

from deepchem.models.tensorflow_models.fcnet import TensorflowMultiTaskClassifier

class TestTensorflowOverfitAPI(TestAPI):

  """

  Test tensorflow models can overfit simple datasets.

  """

  def test_classification_overfit(self):

    """Test that data associated with a tasks stays associated with it."""

    tasks = ["task0"]

    task_types = {task: "classification" for task in tasks}

    n_samples = 10

    n_features = 3

    n_tasks = len(tasks)

    # Generate dummy dataset

    ids = np.arange(n_samples)

    #X = np.random.rand(n_samples, n_features)

    X = np.ones((n_samples, n_features))

    #y = np.random.randint(2, size=(n_samples, n_tasks))

    y = np.zeros((n_samples, n_tasks))

    w = np.ones((n_samples, n_tasks))

    dataset = Dataset.from_numpy(self.train_dir, tasks, X, y, w, ids)

    model_params = {

      "batch_size": 2,

      "num_classification_tasks": 1,

      "num_features": n_features,

      "layer_sizes": [1024],

      "weight_init_stddevs": [.01],

      "bias_init_consts": [0.],

      "dropouts": [.5],

      "num_classes": 2,

      "nb_epoch": 100,

      "penalty": 0.0,

      "optimizer": "sgd",

      "learning_rate": .0003,

      "data_shape": dataset.get_data_shape()

    }

    verbosity = "high"

    classification_metric = Metric(metrics.roc_auc_score)

    model = TensorflowModel(

        tasks, task_types, model_params, self.model_dir,

        tf_class=TensorflowMultiTaskClassifier,

        verbosity=verbosity)

    # Fit trained model

    model.fit(dataset)

    model.save()

    # Eval model on train

    transformers = []

    evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity)

    with tempfile.NamedTemporaryFile() as csv_out:

      with tempfile.NamedTemporaryFile() as stats_out:

        multitask_scores = evaluator.compute_model_performance(

            [classification_metric], csv_out.name, stats_out)

    print("multitask_scores")

    print(multitask_scores)

    assert multitask_scores[classification_metric.name] > .9