def apply_test_data():
data, test_data, cat_col_names, num_col_names = data_load()
bsize = 2500 * 3 * 2 * 2
data_config = DataConfig(target=["target"],
continuous_cols=num_col_names,
categorical_cols=cat_col_names,
num_workers=4)
trainer_config = TrainerConfig(auto_lr_find=True,
batch_size=bsize,
max_epochs=100,
gpus=1)
optimizer_config = OptimizerConfig()
model_config = TabNetModelConfig(task="classification",
learning_rate=1e-3 * bsize / 1024,
n_d=24,
n_a=24,
n_steps=5,
gamma=1.3)
tabular_mode = TabularModel(data_config=data_config,
optimizer_config=optimizer_config,
model_config=model_config,
trainer_config=trainer_config)
for i in range(10):
diri = f"Analysis/basic_tabnet_rep{i}"
tabular_mode.load_from_checkpoint(dir=diri)
def main():
# Generate Synthetic Data
data, test_data, cat_col_names, num_col_names = data_load()
bsize = 512*5
# ##########Define the Configs############
data_config = DataConfig(
target=["target"],
continuous_cols=num_col_names,
categorical_cols=cat_col_names,
num_workers=4
)
trainer_config = TrainerConfig(
auto_lr_find=True,
batch_size=bsize,
max_epochs=100,
gpus=1
)
optimizer_config = OptimizerConfig()
# model_config = TabNetModelConfig(
# task="classification",
# learning_rate=1e-3*bsize/1024,
# n_d=16,
# n_a=16,
# n_steps=5,
# gamma=1.3
# )
model_config = NodeConfig(
task="classification",
num_layers=2, # Number of Dense Layers
num_trees=512, # Number of Trees in each layer
depth=4, # Depth of each Tree
embed_categorical=True,
# If True, will use a learned embedding, else it will use LeaveOneOutEncoding for categorical columns
learning_rate=1e-3,
additional_tree_output_dim=10
)
# Training the Model
# tabular_mode.fit(train=train, validation=val)
# # Evaluating the Model
# # #Loss and Metrics on New Data¶
# result = tabular_mode.evaluate(test)
cv = StratifiedKFold(n_splits=10, shuffle=True)
res_pred = []
res_test = []
train, val = train_test_split(data, random_state=42)
tabular_mode = TabularModel(
data_config=data_config,
optimizer_config=optimizer_config,
model_config=model_config,
trainer_config=trainer_config
)
weighted_loss = get_class_weighted_cross_entropy(train["target"].values.ravel(), mu=0.1)
# Training the Model
tabular_mode.fit(train=train, validation=val, max_epochs=100, loss=weighted_loss)
# tabular_mode.save_model(f"Analysis/basic_node_rep{i}")
ns = 20000
nrep = int(test_data.shape[0]/ns)
nlist = []
for i in range(nrep):
pp = tabular_mode.predict(test_data.iloc[np.arange(ns * i, ns * (i + 1))])
nlist.append(pp)
# #New Predictions as DataFrame
pred_tot = pd.concat(nlist).sort_index()
# print_metrics(data['target'], pred_tot["prediction"], tag="Holdout")
# pred_df = pd.concat([res_testi.loc[:, ["0_probability"]] for res_testi in res_test], axis=1).apply(np.mean, axis=1)
# pred_df2 = pred_df.map(lambda x: 1 if x>0.5 else 0)
sample_submisson = pd.read_csv("Data/sample_submission.csv")
sample_submisson["target"] = pred_tot.prediction.values
sample_submisson.to_csv("Analysis/submission_2_node.csv", index=False)
print(confusion_matrix(data['target'], pred_tot["prediction"]))
def main_64():
# Generate Synthetic Data
global train
data, test_data, cat_col_names, num_col_names = data_load()
bsize = 2500*3*2*2
# ##########Define the Configs############
data_config = DataConfig(
target=["target"],
continuous_cols=num_col_names,
categorical_cols=cat_col_names,
num_workers=4
)
trainer_config = TrainerConfig(
auto_lr_find=True,
batch_size=bsize,
max_epochs=100,
gpus=1
)
optimizer_config = OptimizerConfig()
model_config = TabNetModelConfig(
task="classification",
learning_rate=1e-3*bsize/1024,
n_d=64,
n_a=64,
n_steps=5,
gamma=1.3
)
# Training the Model
# tabular_mode.fit(train=train, validation=val)
# # Evaluating the Model
# # #Loss and Metrics on New Data¶
# result = tabular_mode.evaluate(test)
cv = StratifiedKFold(n_splits=10, shuffle=True)
res_pred = []
res_test = []
for i, (train_idx, test_idx) in enumerate(cv.split(X=data, y=data.target.values)):
train, test = data.iloc[train_idx], data.iloc[test_idx]
train, val = train_test_split(train, random_state=42)
tabular_mode = TabularModel(
data_config=data_config,
optimizer_config=optimizer_config,
model_config=model_config,
trainer_config=trainer_config
)
weighted_loss = get_class_weighted_cross_entropy(train["target"].values.ravel(), mu=0.1)
# Training the Model
tabular_mode.fit(train=train, validation=val, max_epochs=100, loss=weighted_loss)
pred_df = tabular_mode.predict(test).loc[:, ["prediction"]]
res_pred.append(pred_df)
tabular_mode.save_model(f"Analysis/basic_tabnet_rep{i}")
pred = tabular_mode.predict(test_data)
res_test.append(pred)
# #New Predictions as DataFrame
pred_tot = pd.concat(res_pred).sort_index()
print_metrics(data['target'], pred_tot["prediction"], tag="Holdout")
pred_df = pd.concat([res_testi.loc[:, ["0_probability"]] for res_testi in res_test], axis=1).apply(np.mean, axis=1)
pred_df2 = pred_df.map(lambda x: 1 if x>0.5 else 0)
sample_submisson = pd.read_csv("Data/sample_submission.csv")
sample_submisson["target"] = pred_df2.values
sample_submisson.to_csv("Analysis/submission_2.csv", index=False)
print(confusion_matrix(data['target'], pred_tot["prediction"]))