def borealsites_predictions(
data_dir="OneDrive\Dokumente\Sc_Master\Masterthesis\Project\DomAdapt"):
res = pd.read_csv(
os.path.join(
data_dir,
r"python\outputs\models\mlp0\noPool\relu\selected_results.csv"))
dimensions = [7]
for hdim in literal_eval(res["hiddensize"].item()):
dimensions.append(hdim)
dimensions.append(1)
X, Y = preprocessing.get_borealsites(year="both")
X = torch.tensor(X).type(dtype=torch.float)
val_errors_mlp = {"rmse": [], "mae": []}
for i in range(5):
model = models.MLP(dimensions, nn.ReLU)
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp0\\noPool\\relu\model{i}.pth")))
preds = model(X)
val_errors_mlp["rmse"].append(utils.rmse(Y, preds.detach().numpy()))
val_errors_mlp["mae"].append(
metrics.mean_absolute_error(Y,
preds.detach().numpy()))
val_errors_mlp = [
np.mean(val_errors_mlp["rmse"]),
np.mean(val_errors_mlp["mae"])
]
preles_preds = preprocessing.get_borealsites(year="both", preles=True)
val_errors_preles = [
utils.rmse(Y, preles_preds),
metrics.mean_absolute_error(Y, preles_preds)
]
prediction_errors = {
"mlp_prediction_errors": val_errors_mlp,
"preles_prediction_errors": val_errors_preles
}
return (prediction_errors)
df_new = pd.DataFrame({"PAR": X_test["PAR"].mean(),
"TAir": np.arange(X_test["TAir"].min(), X_test["TAir"].max(), step=0.01),
"VPD": X_test["VPD"].mean(),
"Precip": X_test["Precip"].mean(),
"fAPAR": X_test["fAPAR"].mean(),
"DOY_sin": X_test["DOY_sin"].mean(),
"DOY_cos": X_test["DOY_cos"].mean()})
df_new.to_csv(os.path.join(data_dir, "data\post_analysis\df1.csv"), sep=",")
df = df_new.to_numpy()
#%%
preds_reference = []
for i in range(5):
hparams, model_design, X, Y, X_test, Y_test = finetuning.settings("mlp", 10, None, data_dir)
model = models.MLP(model_design["dimensions"], model_design["activation"])
model.load_state_dict(torch.load(os.path.join(data_dir, f"python\outputs\models\mlp0\\relu\model{i}.pth")))
x = torch.tensor(df).type(dtype=torch.float)
preds_reference.append(model(x).detach().numpy())
preds_finetuned= []
for i in range(5):
hparams, model_design, X, Y, X_test, Y_test = finetuning.settings("mlp", 10, None, data_dir)
model = models.MLP(model_design["dimensions"], model_design["activation"])
model.load_state_dict(torch.load(os.path.join(data_dir, f"python\outputs\models\mlp10\\nodropout\sims_frac100\\tuned\setting1\model{i}.pth")))
x = torch.tensor(df).type(dtype=torch.float)
preds_finetuned.append(model(x).detach().numpy())
#%%
m = np.mean(np.transpose(np.array(preds_reference).squeeze(2)), axis=1)
q1 = np.quantile(np.transpose(np.array(preds_reference).squeeze(2)), 0.05, axis=1)
def train_model_CV(hparams,
model_design,
X,
Y,
eval_set,
dropout_prob,
dropout,
data_dir,
save,
splits=5):
"""
"""
epochs = hparams["epochs"]
featuresize = model_design["featuresize"]
kf = KFold(n_splits=splits, shuffle=False)
kf.get_n_splits(X)
rmse_train = np.zeros((splits, epochs))
rmse_val = np.zeros((splits, epochs))
mae_train = np.zeros((splits, epochs))
mae_val = np.zeros((splits, epochs))
# z-score data
#X_mean, X_std = np.mean(X), np.std(X)
#X = utils.minmax_scaler(X)
if not eval_set is None:
print("Test set used for model evaluation")
Xt_test = eval_set["X_test"]
yt_test = eval_set["Y_test"]
#Xt_test= utils.minmax_scaler(Xt_test, scaling = [X_mean, X_std])
yt_test = torch.tensor(yt_test).type(dtype=torch.float)
Xt_test = torch.tensor(Xt_test).type(dtype=torch.float)
#yt_tests = []
i = 0
performance = []
y_tests = []
y_preds = []
for train_index, test_index in kf.split(X):
X_train, X_test = X[train_index], X[test_index]
y_train, y_test = Y[train_index], Y[test_index]
X_test = torch.tensor(X_test).type(dtype=torch.float)
y_test = torch.tensor(y_test).type(dtype=torch.float)
X_train = torch.tensor(X_train).type(dtype=torch.float)
y_train = torch.tensor(y_train).type(dtype=torch.float)
if featuresize is None:
model = models.MLP(model_design["dimensions"],
model_design["activation"])
else:
model = models.MLPmod(featuresize, model_design["dimensions"],
model_design["activation"], dropout_prob,
dropout)
optimizer = optim.Adam(model.parameters(), lr=hparams["learningrate"])
criterion = nn.MSELoss()
for epoch in range(epochs):
# Training
model.train()
x, y = utils.create_batches(X_train, y_train, hparams["batchsize"],
hparams["history"])
x = torch.tensor(x).type(dtype=torch.float)
y = torch.tensor(y).type(dtype=torch.float)
output = model(x)
# Compute training loss
loss = criterion(output, y)
optimizer.zero_grad()
loss.backward()
optimizer.step()
# Evaluate current model at test set
model.eval()
with torch.no_grad():
pred_train = model(X_train)
if eval_set is None:
pred_test = model(X_test)
rmse_train[i, epoch] = utils.rmse(y_train, pred_train)
rmse_val[i, epoch] = utils.rmse(y_test, pred_test)
mae_train[i, epoch] = metrics.mean_absolute_error(
y_train, pred_train)
mae_val[i, epoch] = metrics.mean_absolute_error(
y_test, pred_test)
else:
pred_test = model(Xt_test)
rmse_train[i, epoch] = utils.rmse(y_train, pred_train)
rmse_val[i, epoch] = utils.rmse(yt_test, pred_test)
mae_train[i, epoch] = metrics.mean_absolute_error(
y_train, pred_train)
mae_val[i, epoch] = metrics.mean_absolute_error(
yt_test, pred_test)
if save:
if epoch % 1000 == 0:
print("Epoch", epoch, ": Saving model to path.")
torch.save(model.state_dict(),
os.path.join(data_dir, f"model{i}.pth"))
# Predict with fitted model
with torch.no_grad():
preds_train = model(X_train)
if eval_set is None:
preds_test = model(X_test)
performance.append([
utils.rmse(y_train, preds_train),
utils.rmse(y_test, preds_test),
metrics.mean_absolute_error(y_train, preds_train.numpy()),
metrics.mean_absolute_error(y_test, preds_test.numpy())
])
else:
preds_test = model(Xt_test)
performance.append([
utils.rmse(y_train, preds_train),
utils.rmse(yt_test, preds_test),
metrics.mean_absolute_error(y_train, preds_train.numpy()),
metrics.mean_absolute_error(yt_test, preds_test.numpy())
])
if eval_set is None:
y_tests.append(y_test.numpy())
else:
y_tests.append(yt_test.numpy())
y_preds.append(preds_test.numpy())
i += 1
running_losses = {
"rmse_train": rmse_train,
"mae_train": mae_train,
"rmse_val": rmse_val,
"mae_val": mae_val
}
return (running_losses, performance, y_tests, y_preds)
def featureExtractorD(
model,
typ,
epochs,
simsfrac,
splits=5,
data_dir="OneDrive\Dokumente\Sc_Master\Masterthesis\Project\DomAdapt"):
hparams, model_design, X, Y, X_test, Y_test = settings(
"mlp", None, data_dir)
hparams_add, model_design_add, X, Y, X_test, Y_test = settings("mlp",
epochs,
data_dir,
sims=False)
X = torch.tensor(X).type(dtype=torch.float)
X_test = torch.tensor(X_test).type(dtype=torch.float)
errors = []
preds_tests = []
for i in range(splits):
# Load pretrained model
model = models.MLPmod(model_design["featuresize"],
model_design["dimensions"],
model_design["activation"])
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp{typ}\\nodropout\sims_frac{simsfrac}\model{i}.pth"
)))
# modify classifier
model.classifier = nn.Sequential(
*list(model.classifier.children())[:-1])
# extract features
out_train = model(X).detach().numpy()
out_test = model(X_test).detach().numpy()
# specify dimensions of model to train (architecture 2)
model_design_add["dimensions"].insert(0, out_train.shape[1])
# Train mlp with extracted features as input, predicting Y.
running_losses, pred_test = train_model(hparams_add, model_design_add,
out_train, Y, out_test, Y_test,
i)
# Evaluate model (reload it.)
model = models.MLP(model_design_add["dimensions"],
model_design_add["activation"])
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp7\\nodropout\sims_frac30\\tuned\setting2\model{i}.pth"
)))
preds_test = model(
torch.tensor(out_test).type(dtype=torch.float)).detach().numpy()
preds_train = model(
torch.tensor(out_train).type(dtype=torch.float)).detach().numpy()
errors.append([
utils.rmse(Y, preds_train),
utils.rmse(Y_test, preds_test),
metrics.mean_absolute_error(Y, preds_train),
metrics.mean_absolute_error(Y_test, preds_test)
])
preds_tests.append(preds_test)
return (running_losses, errors, preds_test)
def train_model(
hparams_add,
model_design_add,
X,
Y,
X_test,
Y_test,
i,
data_dir="OneDrive\Dokumente\Sc_Master\Masterthesis\Project\DomAdapt"):
epochs = hparams_add["epochs"]
rmse_train = np.zeros((epochs))
rmse_val = np.zeros((epochs))
mae_train = np.zeros((epochs))
mae_val = np.zeros((epochs))
# Standardize X and X_test together!!
#mu = np.concatenate((X, X_test), 0).mean()
#sigma = np.concatenate((X, X_test), 0).std()
#X = utils.minmax_scaler(X, [mu, sigma])
#X_test = utils.minmax_scaler(X_test, [mu, sigma])
X_test = torch.tensor(X_test).type(dtype=torch.float)
y_test = torch.tensor(Y_test).type(dtype=torch.float)
X_train = torch.tensor(X).type(dtype=torch.float)
y_train = torch.tensor(Y).type(dtype=torch.float)
model_design_add["dimensions"].insert(0, X.shape[1])
model = models.MLP(model_design_add["dimensions"],
model_design_add["activation"])
optimizer = optim.Adam(model.parameters(), lr=hparams_add["learningrate"])
criterion = nn.MSELoss()
for epoch in range(epochs):
# Training
model.train()
x, y = utils.create_batches(X_train, y_train, hparams_add["batchsize"],
hparams_add["history"])
x = torch.tensor(x).type(dtype=torch.float)
y = torch.tensor(y).type(dtype=torch.float)
output = model(x)
# Compute training loss
loss = criterion(output, y)
optimizer.zero_grad()
loss.backward()
optimizer.step()
# Evaluate current model at test set
model.eval()
with torch.no_grad():
pred_train = model(X_train)
pred_test = model(X_test)
rmse_train[epoch] = utils.rmse(y_train, pred_train)
rmse_val[epoch] = utils.rmse(y_test, pred_test)
mae_train[epoch] = metrics.mean_absolute_error(y_train, pred_train)
mae_val[epoch] = metrics.mean_absolute_error(y_test, pred_test)
torch.save(
model.state_dict(),
os.path.join(
data_dir,
f"python\outputs\models\mlp7\\nodropout\sims_frac30\\tuned\setting2\model{i}.pth"
))
running_losses = {
"rmse_train": rmse_train,
"mae_train": mae_train,
"rmse_val": rmse_val,
"mae_val": mae_val
}
return running_losses, pred_test
import pandas as pd
import setup.models as models
import torch.nn as nn
import torch
from ast import literal_eval
#%%
data_dir = "OneDrive\Dokumente\Sc_Master\Masterthesis\Project\DomAdapt"
rets_mlp = pd.read_csv(
os.path.join(
data_dir,
r"python\outputs\grid_search\observations\mlp\grid_search_results_mlp2.csv"
))
rets_mlp = rets_mlp[(rets_mlp.nlayers == 3)].reset_index()
bm = rets_mlp.iloc[rets_mlp['mae_val'].idxmin()].to_dict()
model = models.MLP([7, 128, 16, 256, 1], nn.ReLU)
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
r"python\outputs\models\mlp4\nodropout\sims_frac100\model0.pth")))
#%%
for child in model.children():
print(child)
for name, parameter in child.named_parameters():
print(name)
print(parameter)
if not name in ["hidden3.weight", "hidden3.bias"]:
print("disable backprob for", name)
parameter.requires_grad = False
#%%
def featureExtractorC(
typ,
epochs,
simsfrac,
dummies=False,
sparse=None,
classifier="ols",
years=[2001, 2002, 2003, 2004, 2005, 2006, 2007],
random_days=None,
splits=5,
data_dir="OneDrive\Dokumente\Sc_Master\Masterthesis\Project\DomAdapt"):
hparams, model_design, X, Y, X_test, Y_test = settings(
typ,
epochs,
data_dir,
dummies,
sparse,
years=years,
random_days=random_days)
X = torch.tensor(X).type(dtype=torch.float)
X_test = torch.tensor(X_test).type(dtype=torch.float)
predictions_train = []
predictions_test = []
for i in range(splits):
if ((typ == 4) | (typ == 9) | (typ == 10) | (typ == 11) | (typ == 12) |
(typ == 13) | (typ == 14)):
model = models.MLP(model_design["dimensions"],
model_design["activation"])
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp{typ}\\nodropout\sims_frac{simsfrac}\model{i}.pth"
)))
elif typ == 7:
print("load for model", typ)
model = models.MLPmod(model_design["featuresize"],
model_design["dimensions"],
model_design["activation"])
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp{typ}\\nodropout\sims_frac{simsfrac}\model{i}.pth"
)))
elif typ == 5:
print("load for model", typ)
model = models.MLPmod(model_design["featuresize"],
model_design["dimensions"],
model_design["activation"])
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp{typ}\\nodropout\dummies\sims_frac{simsfrac}\model{i}.pth"
)))
if ((typ == 4) | (typ == 9) | (typ == 10) | (typ == 11) | (typ == 12) |
(typ == 13) | (typ == 14)):
model = model[:-1]
else:
model.classifier = nn.Sequential(
*list(model.classifier.children())
[:-1]) # Remove Final layer and activation.
out_train = model(X).detach().numpy()
out_train = sm.add_constant(out_train) # Add intercept.
out_test = model(X_test).detach().numpy()
out_test = sm.add_constant(out_test) # Add intercept.
if classifier == "ols":
extractor = sm.OLS(Y, out_train)
results = extractor.fit()
predictions_train.append(np.expand_dims(results.predict(), axis=1))
predictions_test.append(
np.expand_dims(results.predict(out_test), axis=1))
elif classifier == "glm":
print("Fitting glm with Inverse Gaussian family and log-Link.")
extractor = sm.GLM(Y,
out_train,
family=sm.families.InverseGaussian(
sm.families.links.log()))
results = extractor.fit()
predictions_train.append(np.expand_dims(results.predict(), axis=1))
predictions_test.append(
np.expand_dims(results.predict(out_test), axis=1))
elif classifier == "nnls":
theta = np.expand_dims(nnls(out_train, Y[:, 0])[0], axis=1)
predictions_train.append(np.dot(out_train, theta))
predictions_test.append(np.dot(out_test, theta))
else:
print("Don't know classifier.")
mae_train = [
metrics.mean_absolute_error(Y, sublist)
for sublist in predictions_train
]
mae_val = [
metrics.mean_absolute_error(Y_test, sublist)
for sublist in predictions_test
]
rmse_train = [utils.rmse(Y, sublist) for sublist in predictions_train]
rmse_val = [utils.rmse(Y_test, sublist) for sublist in predictions_test]
errors = [rmse_train, rmse_val, mae_train, mae_val]
return predictions_test, errors
def featureExtractorA(
typ,
epochs,
simsfrac,
dummies,
sparse=None,
years=[2001, 2002, 2003, 2004, 2005, 2006, 2007],
random_days=None,
splits=5,
data_dir="OneDrive\Dokumente\Sc_Master\Masterthesis\Project\DomAdapt"):
hparams, model_design, X, Y, X_test, Y_test = settings(
typ,
epochs,
data_dir,
dummies,
sparse,
years=years,
random_days=random_days)
if ((typ == 4) | (typ == 9) | (typ == 10) | (typ == 12) | (typ == 13) |
(typ == 14)):
model_design["featuresize"] = None
X = torch.tensor(X).type(dtype=torch.float)
X_test = torch.tensor(X_test).type(dtype=torch.float)
predictions = []
mae_train = []
rmse_train = []
mae_val = []
rmse_val = []
for i in range(splits):
if ((typ == 4) | (typ == 9) | (typ == 10) | (typ == 11) | (typ == 12) |
(typ == 13) | (typ == 14)):
model = models.MLP(model_design["dimensions"],
model_design["activation"])
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp{typ}\\nodropout\sims_frac{simsfrac}\model{i}.pth"
)))
elif typ == 7:
print("load for model", typ)
model = models.MLPmod(model_design["featuresize"],
model_design["dimensions"],
model_design["activation"])
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp{typ}\\nodropout\sims_frac{simsfrac}\model{i}.pth"
)))
elif typ == 5:
print("load for model", typ)
model = models.MLPmod(model_design["featuresize"],
model_design["dimensions"],
model_design["activation"])
model.load_state_dict(
torch.load(
os.path.join(
data_dir,
f"python\outputs\models\mlp{typ}\\nodropout\dummies\sims_frac{simsfrac}\model{i}.pth"
)))
preds_test = model(X_test).detach().numpy()
preds_train = model(X).detach().numpy()
mae_val.append(metrics.mean_absolute_error(Y_test, preds_test))
rmse_val.append(utils.rmse(Y_test, preds_test))
mae_train.append(metrics.mean_absolute_error(Y, preds_train))
rmse_train.append(utils.rmse(Y, preds_train))
predictions.append(preds_test)
errors = [rmse_train, rmse_val, mae_train, mae_val]
return predictions, errors
x, target = utils.create_batches(X, Y, 128, 0)
#x_test, target_test = utils.create_batches(X, Y, 128, 0)
x = torch.tensor(x).type(dtype=torch.float)
target = torch.tensor(target).type(dtype=torch.float)
#x_test = torch.tensor(x_test).type(dtype=torch.float)
#target_test = torch.tensor(target_test).type(dtype=torch.float)
#%%
hiddensize = [16, 32, 64, 128, 256]
dimensions = [X.shape[1]]
for layer in range(2):
# randomly pick hiddensize from hiddensize list
dimensions.append(random.choice(hiddensize))
dimensions.append(Y.shape[1])
model = models.MLP(dimensions, nn.ReLU)
out = model(x)
#%%
x_test, target_test = utils.create_inout_sequences(X, Y, 64, 10, model="cnn")
rmse_train = []
rmse_test = []
model = models.ConvN([X.shape[1], 32, 1], [X.shape[1], 14], 2, 10, nn.ReLU)
optimizer = optim.Adam(model.parameters(), lr=0.01)
criterion = nn.MSELoss()
x_train, y_train = utils.create_inout_sequences(X, Y, "full", 10, model="cnn")
datadir = "OneDrive\Dokumente\Sc_Master\Masterthesis\Project\DomAdapt"
X, Y = preprocessing.get_splits(sites=['hyytiala'],
years=[2008],
datadir=os.path.join(datadir, "data"),
dataset="profound",
simulations=None)
#%%
gridsearch_results = pd.read_csv(
os.path.join(
datadir,
f"python\outputs\grid_search\mlp\grid_search_results_mlp1.csv"))
setup = gridsearch_results.iloc[
gridsearch_results['mae_val'].idxmin()].to_dict()
model = models.MLP([7, 64, 64, 16, 1], nn.ReLU)
model.load_state_dict(
torch.load(os.path.join(datadir,
f"python\outputs\models\mlp2\model0.pth")))
X_scaled = utils.minmax_scaler(X)
X_scaled = torch.tensor(X_scaled).type(dtype=torch.float)
y_preds = model(X_scaled).detach().numpy()
plt.plot(Y)
plt.plot(y_preds)
metrics.mean_absolute_error(Y, y_preds)
#%%
model = models.MLP([7, 64, 64, 16, 1], nn.ReLU)
def training_CV(hparams, model_design, X, Y, feature_extraction, eval_set, data_dir,
save, sparse=None, dropout_prob = 0.0, dropout = False, splits = 5):
"""
"""
epochs = hparams["epochs"]
featuresize = model_design["featuresize"]
kf = KFold(n_splits=splits, shuffle = False)
kf.get_n_splits(X)
rmse_train = np.zeros((splits, epochs))
rmse_val = np.zeros((splits, epochs))
mae_train = np.zeros((splits, epochs))
mae_val = np.zeros((splits, epochs))
# z-score data
#X_mean, X_std = np.mean(X), np.std(X)
#X = utils.minmax_scaler(X)
if not eval_set is None:
print("Test set used for model evaluation")
Xt_test = eval_set["X_test"]
#Xt_test= utils.minmax_scaler(Xt_test, scaling = [X_mean, X_std])
yt_test = eval_set["Y_test"]
yt_test = torch.tensor(yt_test).type(dtype=torch.float)
Xt_test = torch.tensor(Xt_test).type(dtype=torch.float)
yt_tests = []
i = 0
performance = []
y_tests = []
y_preds = []
for train_index, test_index in kf.split(X):
X_train, X_test = X[train_index], X[test_index]
y_train, y_test = Y[train_index], Y[test_index]
X_test = torch.tensor(X_test).type(dtype=torch.float)
y_test = torch.tensor(y_test).type(dtype=torch.float)
X_train = torch.tensor(X_train).type(dtype=torch.float)
y_train = torch.tensor(y_train).type(dtype=torch.float)
if not featuresize is None:
if isinstance(model_design, dict):
print("Loading pretrained Model.")
model = models.MLPmod(model_design["featuresize"], model_design["dimensions"], model_design["activation"], dropout_prob, dropout)
model.load_state_dict(torch.load(os.path.join(data_dir, f"model{i}.pth")))
else:
model = model_design
else:
model = models.MLP(model_design["dimensions"], model_design["activation"])
model.load_state_dict(torch.load(os.path.join(data_dir, f"model{i}.pth")))
model.eval()
if not feature_extraction is None:
print("Freezing all weights.")
if featuresize is None:
try:
print("backpropagation of third layer parameters.")
model.hidden3.weight.requires_grad = False
model.hidden3.bias.requires_grad = False
except:
print("backpropagation of first layer parameters.")
model.hidden1.weight.requires_grad = False
model.hidden1.bias.requires_grad = False
else:
for child in model.children():
print("Entering child node")
for name, parameter in child.named_parameters():
#print(name)
if not name in feature_extraction:
print("disable backprob for", name)
parameter.requires_grad = False
#else:
# parameter.requires_grad = False
criterion = nn.MSELoss()
optimizer = optim.Adam(model.parameters(), lr = hparams["learningrate"])
for epoch in range(epochs):
# Training
model.train()
x, y = utils.create_batches(X_train, y_train, hparams["batchsize"], hparams["history"])
x = torch.tensor(x).type(dtype=torch.float)
y = torch.tensor(y).type(dtype=torch.float)
output = model(x)
# Compute training loss
loss = criterion(output, y)
optimizer.zero_grad()
loss.backward()
optimizer.step()
# Evaluate current model at test set
model.eval()
with torch.no_grad():
pred_train = model(X_train)
if eval_set is None:
pred_test = model(X_test)
rmse_train[i, epoch] = utils.rmse(y_train, pred_train)
rmse_val[i, epoch] = utils.rmse(y_test, pred_test)
mae_train[i, epoch] = metrics.mean_absolute_error(y_train, pred_train)
mae_val[i, epoch] = metrics.mean_absolute_error(y_test, pred_test)
else:
pred_test = model(Xt_test)
rmse_train[i, epoch] = utils.rmse(y_train, pred_train)
rmse_val[i, epoch] = utils.rmse(yt_test, pred_test)
mae_train[i, epoch] = metrics.mean_absolute_error(y_train, pred_train)
mae_val[i, epoch] = metrics.mean_absolute_error(yt_test, pred_test)
# Predict with fitted model
with torch.no_grad():
preds_train = model(X_train)
if eval_set is None:
preds_test = model(X_test)
performance.append([utils.rmse(y_train, preds_train),
utils.rmse(y_test, preds_test),
metrics.mean_absolute_error(y_train, preds_train.numpy()),
metrics.mean_absolute_error(y_test, preds_test.numpy())])
else:
preds_test = model(Xt_test)
performance.append([utils.rmse(y_train, preds_train),
utils.rmse(yt_test, preds_test),
metrics.mean_absolute_error(y_train, preds_train.numpy()),
metrics.mean_absolute_error(yt_test, preds_test.numpy())])
if save:
if not feature_extraction is None:
if not sparse is None:
torch.save(model.state_dict(), os.path.join(data_dir, f"tuned/setting1/sparse//{sparse}/model{i}.pth"))
else:
torch.save(model.state_dict(), os.path.join(data_dir, f"tuned/setting1/model{i}.pth"))
else:
if not sparse is None:
torch.save(model.state_dict(), os.path.join(data_dir, f"tuned/setting0/sparse//{sparse}/model{i}.pth"))
else:
torch.save(model.state_dict(), os.path.join(data_dir, f"tuned/setting0/model{i}.pth"))
y_tests.append(y_test.numpy())
y_preds.append(preds_test.numpy())
i += 1
running_losses = {"rmse_train":rmse_train, "mae_train":mae_train, "rmse_val":rmse_val, "mae_val":mae_val}
if eval_set is None:
return(running_losses, performance, y_tests, y_preds)
else:
return(running_losses, performance, yt_tests, y_preds)