Commit 95d32dfc authored by hsjang001205's avatar hsjang001205
Browse files

DAG_reload

parent e5827cc9

deepchem/models/layers.py

+119 −269

File changed.

Preview size limit exceeded, changes collapsed.

deepchem/models/tests/test_reload.py

+164 −153
Original line number Diff line number Diff line
@@ -126,15 +126,14 @@ def test_multitaskclassification_reload(): 


  classification_metric = dc.metrics.Metric(dc.metrics.accuracy_score)

  model_dir = tempfile.mkdtemp()

  model = dc.models.MultitaskClassifier(n_tasks,

  model = dc.models.MultitaskClassifier(

      n_tasks,

      n_features,

      dropouts=[0.],

      weight_init_stddevs=[.1],

      batch_size=n_samples,

      optimizer=dc.models.optimizers.Adam(

                                            learning_rate=0.0003,

                                            beta1=0.9,

                                            beta2=0.999),

          learning_rate=0.0003, beta1=0.9, beta2=0.999),

      model_dir=model_dir)



  # Fit trained model
@@ -147,9 +146,8 @@ def test_multitaskclassification_reload(): 
      dropouts=[0.],

      weight_init_stddevs=[.1],

      batch_size=n_samples,

      optimizer=dc.models.optimizers.Adam(learning_rate=0.0003,

                                          beta1=0.9,

                                          beta2=0.999),

      optimizer=dc.models.optimizers.Adam(

          learning_rate=0.0003, beta1=0.9, beta2=0.999),

      model_dir=model_dir)

  reloaded_model.restore()
@@ -182,7 +180,8 @@ def test_residual_classification_reload(): 


  classification_metric = dc.metrics.Metric(dc.metrics.accuracy_score)

  model_dir = tempfile.mkdtemp()

  model = dc.models.MultitaskClassifier(n_tasks,

  model = dc.models.MultitaskClassifier(

      n_tasks,

      n_features,

      layer_sizes=[20] * 10,

      dropouts=0.0,
@@ -198,7 +197,8 @@ def test_residual_classification_reload(): 
  assert scores[classification_metric.name] > .9



  # Reload trained model

  reloaded_model = dc.models.MultitaskClassifier(n_tasks,

  reloaded_model = dc.models.MultitaskClassifier(

      n_tasks,

      n_features,

      layer_sizes=[20] * 10,

      dropouts=0.0,
@@ -234,10 +234,11 @@ def test_robust_multitask_classification_reload(): 
  w = np.ones((n_samples, n_tasks))

  dataset = dc.data.NumpyDataset(X, y, w, ids)



  classification_metric = dc.metrics.Metric(dc.metrics.accuracy_score,

                                            task_averager=np.mean)

  classification_metric = dc.metrics.Metric(

      dc.metrics.accuracy_score, task_averager=np.mean)

  model_dir = tempfile.mkdtemp()

  model = dc.models.RobustMultitaskClassifier(n_tasks,

  model = dc.models.RobustMultitaskClassifier(

      n_tasks,

      n_features,

      layer_sizes=[50],

      bypass_layer_sizes=[10],
@@ -255,7 +256,8 @@ def test_robust_multitask_classification_reload(): 
  assert scores[classification_metric.name] > .9



  # Reloaded Trained Model

  reloaded_model = dc.models.RobustMultitaskClassifier(n_tasks,

  reloaded_model = dc.models.RobustMultitaskClassifier(

      n_tasks,

      n_features,

      layer_sizes=[50],

      bypass_layer_sizes=[10],
@@ -295,7 +297,8 @@ def test_robust_multitask_regressor_reload(): 
  regression_metric = dc.metrics.Metric(dc.metrics.mean_squared_error)



  model_dir = tempfile.mkdtemp()

  model = dc.models.RobustMultitaskRegressor(n_tasks,

  model = dc.models.RobustMultitaskRegressor(

      n_tasks,

      n_features,

      layer_sizes=[50],

      bypass_layer_sizes=[10],
@@ -313,7 +316,8 @@ def test_robust_multitask_regressor_reload(): 
  assert scores[regression_metric.name] < .1



  # Reload trained model

  reloaded_model = dc.models.RobustMultitaskRegressor(n_tasks,

  reloaded_model = dc.models.RobustMultitaskRegressor(

      n_tasks,

      n_features,

      layer_sizes=[50],

      bypass_layer_sizes=[10],
@@ -353,10 +357,11 @@ def test_IRV_multitask_classification_reload(): 
  IRV_transformer = dc.trans.IRVTransformer(5, n_tasks, dataset)

  dataset_trans = IRV_transformer.transform(dataset)



  classification_metric = dc.metrics.Metric(dc.metrics.accuracy_score,

                                            task_averager=np.mean)

  classification_metric = dc.metrics.Metric(

      dc.metrics.accuracy_score, task_averager=np.mean)

  model_dir = tempfile.mkdtemp()

  model = dc.models.MultitaskIRVClassifier(n_tasks,

  model = dc.models.MultitaskIRVClassifier(

      n_tasks,

      K=5,

      learning_rate=0.01,

      batch_size=n_samples,
@@ -370,7 +375,8 @@ def test_IRV_multitask_classification_reload(): 
  assert scores[classification_metric.name] > .9



  # Reload Trained Model

  reloaded_model = dc.models.MultitaskIRVClassifier(n_tasks,

  reloaded_model = dc.models.MultitaskIRVClassifier(

      n_tasks,

      K=5,

      learning_rate=0.01,

      batch_size=n_samples,
@@ -406,10 +412,11 @@ def test_progressive_classification_reload(): 


  dataset = dc.data.NumpyDataset(X, y, w, ids)



  classification_metric = dc.metrics.Metric(dc.metrics.accuracy_score,

                                            task_averager=np.mean)

  classification_metric = dc.metrics.Metric(

      dc.metrics.accuracy_score, task_averager=np.mean)

  model_dir = tempfile.mkdtemp()

  model = dc.models.ProgressiveMultitaskClassifier(n_tasks,

  model = dc.models.ProgressiveMultitaskClassifier(

      n_tasks,

      n_features,

      layer_sizes=[50],

      bypass_layer_sizes=[10],
@@ -470,7 +477,8 @@ def test_progressivemultitaskregressor_reload(): 
  regression_metric = dc.metrics.Metric(dc.metrics.mean_squared_error)



  model_dir = tempfile.mkdtemp()

  model = dc.models.ProgressiveMultitaskRegressor(n_tasks,

  model = dc.models.ProgressiveMultitaskRegressor(

      n_tasks,

      n_features,

      layer_sizes=[50],

      bypass_layer_sizes=[10],
@@ -523,14 +531,15 @@ def test_DAG_regression_reload(): 


  # Load mini log-solubility dataset.

  featurizer = dc.feat.ConvMolFeaturizer()

  mols = ["CCCCCC"] * 10

  tasks = ["outcome"]

  mols = ["CC", "CCO", "CC","CCC","CCCCO","CO","CC","CCCCC","CCC","CCCO"]

  n_samples = len(mols)

  X = featurizer(mols)

  y = np.random.rand(n_samples, n_tasks)

  dataset = dc.data.NumpyDataset(X, y)



  regression_metric = dc.metrics.Metric(dc.metrics.pearson_r2_score,

                                        task_averager=np.mean)

  regression_metric = dc.metrics.Metric(

      dc.metrics.pearson_r2_score, task_averager=np.mean)



  n_feat = 75

  batch_size = 10
@@ -538,7 +547,8 @@ def test_DAG_regression_reload(): 
  dataset = transformer.transform(dataset)



  model_dir = tempfile.mkdtemp()

  model = dc.models.DAGModel(n_tasks,

  model = dc.models.DAGModel(

      n_tasks,

      max_atoms=50,

      n_atom_feat=n_feat,

      batch_size=batch_size,
@@ -554,7 +564,8 @@ def test_DAG_regression_reload(): 
  scores = model.evaluate(dataset, [regression_metric])

  assert scores[regression_metric.name] > .1



  reloaded_model = dc.models.DAGModel(n_tasks,

  reloaded_model = dc.models.DAGModel(

      n_tasks,

      max_atoms=50,

      n_atom_feat=n_feat,

      batch_size=batch_size,
@@ -562,7 +573,6 @@ def test_DAG_regression_reload(): 
      use_queue=False,

      mode="regression",

      model_dir=model_dir)



  reloaded_model.restore()

  

  # Check predictions match on random sample
@@ -579,7 +589,6 @@ def test_DAG_regression_reload(): 
  scores = reloaded_model.evaluate(dataset, [regression_metric])

  assert scores[regression_metric.name] > .1





## TODO: THIS IS FAILING!

#def test_weave_classification_reload_alt():

#  """Test weave model can be reloaded."""
@@ -899,7 +908,8 @@ def test_1d_cnn_regression_reload(): 
  regression_metric = dc.metrics.Metric(dc.metrics.mean_squared_error)

  model_dir = tempfile.mkdtemp()



  model = dc.models.CNN(n_tasks,

  model = dc.models.CNN(

      n_tasks,

      n_features,

      dims=1,

      dropouts=0,
@@ -916,7 +926,8 @@ def test_1d_cnn_regression_reload(): 
  assert scores[regression_metric.name] < 0.1



  # Reload trained model

  reloaded_model = dc.models.CNN(n_tasks,

  reloaded_model = dc.models.CNN(

      n_tasks,

      n_features,

      dims=1,

      dropouts=0,
@@ -1005,9 +1016,8 @@ def test_chemception_reload(): 
  img_spec = "engd"

  res = 0.5

  n_tasks = 1

  featurizer = dc.feat.SmilesToImage(img_size=img_size,

                                     img_spec=img_spec,

                                     res=res)

  featurizer = dc.feat.SmilesToImage(

      img_size=img_size, img_spec=img_spec, res=res)



  data_points = 10

  mols = ["CCCCCCCC"] * data_points
@@ -1016,19 +1026,20 @@ def test_chemception_reload(): 
  y = np.random.randint(0, 2, size=(data_points, n_tasks))

  w = np.ones(shape=(data_points, n_tasks))

  dataset = dc.data.NumpyDataset(X, y, w, mols)

  classsification_metric = dc.metrics.Metric(dc.metrics.roc_auc_score,

                                             np.mean,

                                             mode="classification")

  classsification_metric = dc.metrics.Metric(

      dc.metrics.roc_auc_score, np.mean, mode="classification")



  model_dir = tempfile.mkdtemp()

  model = dc.models.ChemCeption(n_tasks=n_tasks,

  model = dc.models.ChemCeption(

      n_tasks=n_tasks,

      img_spec="engd",

      model_dir=model_dir,

      mode="classification")

  model.fit(dataset, nb_epoch=3)



  # Reload Trained Model

  reloaded_model = dc.models.ChemCeption(n_tasks=n_tasks,

  reloaded_model = dc.models.ChemCeption(

      n_tasks=n_tasks,

      img_spec="engd",

      model_dir=model_dir,

      mode="classification")

CRA Git | Maintained and supported by SUSTech CRA and CCSE