Skip to content

Added wandb loggers to tune_ts files #4

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
paavalipopov wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Added wandb loggers to tune_ts files #4

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c6566e6
newer wandb
Apr 27, 2022
a57670c
updated logger
Apr 28, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
68 changes: 60 additions & 8 deletions src/scripts/tune_ts_baseline.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# pylint: disable-all
import argparse
import os

Expand All @@ -15,6 +16,8 @@
from src.settings import LOGS_ROOT, UTCNOW
from src.ts import load_ABIDE1, TSQuantileTransformer

import wandb

sns.set_theme(style="whitegrid", font_scale=2, rc={"figure.figsize": (18, 9)})


Expand Down Expand Up @@ -66,6 +69,9 @@ def on_tune_start(self):

def on_experiment_start(self, exp: "IExperiment"):
super().on_experiment_start(exp)

# init logger
self.wandb_logger: wandb.run = wandb.init(project="tune_ts")
# setup model
clf_type = self._trial.suggest_categorical(
"classifier",
Expand All @@ -76,30 +82,53 @@ def on_experiment_start(self, exp: "IExperiment"):
"RandomForestClassifier",
],
)

if clf_type == "LogisticRegression":
solver = self._trial.suggest_categorical(
"classifier.logistic.solver", ["liblinear", "lbfgs"]
)
decay = self._trial.suggest_loguniform(
"classifier.logistic.C", low=1e-3, high=1e3
)
decay = self._trial.suggest_loguniform("classifier.logistic.C", low=1e-3, high=1e3)
if solver == "liblinear":
penalty = self._trial.suggest_categorical(
"classifier.logistic.penalty", ["l1", "l2"]
)
else:
penalty = "l2"

# configure logger
self.wandb_logger.name = f"{UTCNOW}-{clf_type}"
self.wandb_logger.config.update(
{
"classifier": clf_type,
"solver": solver,
"decay": decay,
"penalty": penalty,
"max_iter": 1000,
}
)

self.classifier = LogisticRegression(
solver=solver, C=decay, penalty=penalty, max_iter=1000
)
elif clf_type == "SGDClassifier":
penalty = self._trial.suggest_categorical(
"classifier.sgd.penalty", ["l1", "l2", "elasticnet"]
)
alpha = self._trial.suggest_loguniform(
"classifier.sgd.alpha", low=1e-4, high=1e-2
alpha = self._trial.suggest_loguniform("classifier.sgd.alpha", low=1e-4, high=1e-2)

# configure logger
self.wandb_logger.name = f"{UTCNOW}-{clf_type}"
self.wandb_logger.config.update(
{
"classifier": clf_type,
"loss": "modified_huber",
"penalty": penalty,
"alpha": alpha,
"max_iter": 1000,
"tol": 1e-3,
}
)

self.classifier = SGDClassifier(
loss="modified_huber",
penalty=penalty,
Expand All @@ -112,6 +141,16 @@ def on_experiment_start(self, exp: "IExperiment"):
n_estimators = self._trial.suggest_int(
"classifier.adaboost.n_estimators", 2, 32, log=True
)

# configure logger
self.wandb_logger.name = f"{UTCNOW}-{clf_type}"
self.wandb_logger.config.update(
{
"classifier": clf_type,
"n_estimators": n_estimators,
}
)

self.classifier = AdaBoostClassifier(n_estimators=n_estimators)
elif clf_type == "RandomForestClassifier":
max_depth = self._trial.suggest_int(
Expand All @@ -120,6 +159,18 @@ def on_experiment_start(self, exp: "IExperiment"):
n_estimators = self._trial.suggest_int(
"classifier.random_forest.n_estimators", 2, 32, log=True
)

# configure logger
self.wandb_logger.name = f"{UTCNOW}-{clf_type}"
self.wandb_logger.config.update(
{
"classifier": clf_type,
"max_depth": max_depth,
"n_estimators": n_estimators,
"max_features": 1,
}
)

self.classifier = RandomForestClassifier(
max_depth=max_depth, n_estimators=n_estimators, max_features=1
)
Expand All @@ -135,24 +186,25 @@ def run_dataset(self) -> None:
num_times=self._num_times,
logpath=f"{self._logdir}/{self._trial.number}.coefs.png",
)
report = get_classification_report(
y_true=y_test, y_pred=y_pred, y_score=y_score, beta=0.5
)
report = get_classification_report(y_true=y_test, y_pred=y_pred, y_score=y_score, beta=0.5)
for stats_type in [0, 1, "macro", "weighted"]:
stats = report.loc[stats_type]
for key, value in stats.items():
if "support" not in key:
self._trial.set_user_attr(f"{key}_{stats_type}", float(value))
self.dataset_metrics = {"score": report["auc"].loc["weighted"]}
self.wandb_logger.log({"score": self.dataset_metrics["score"]})

def on_experiment_end(self, exp: "IExperiment") -> None:
super().on_experiment_end(exp)
# we have only 1 epoch for baselines, so...
self._score = self.experiment_metrics[1]["ABIDE1"]["score"]
self.wandb_logger.finish()

def _objective(self, trial) -> float:
self._trial = trial
self.run()

return self._score

def tune(self, n_trials: int):
Expand Down
59 changes: 46 additions & 13 deletions src/scripts/tune_ts_lstm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# pylint: disable-all
import argparse

from animus import EarlyStoppingCallback, IExperiment
Expand All @@ -16,6 +17,8 @@
from src.settings import LOGS_ROOT, UTCNOW
from src.ts import load_ABIDE1, TSQuantileTransformer

import wandb


class LSTM(nn.Module):
def __init__(
Expand All @@ -28,9 +31,7 @@ def __init__(
super(LSTM, self).__init__()
self.hidden_size = hidden_size
self.bidirectional = bidirectional
self.lstm = nn.LSTM(
hidden_size=hidden_size, bidirectional=bidirectional, **kwargs
)
self.lstm = nn.LSTM(hidden_size=hidden_size, bidirectional=bidirectional, **kwargs)
self.fc = nn.Sequential(
nn.Dropout(p=fc_dropout),
nn.Linear(2 * hidden_size if bidirectional else hidden_size, 1),
Expand Down Expand Up @@ -85,6 +86,9 @@ def on_tune_start(self):
)

def on_experiment_start(self, exp: "IExperiment"):
# init wandb logger
self.wandb_logger: wandb.run = wandb.init(project="tune_lstm", name=f"{UTCNOW}-lstm")

super().on_experiment_start(exp)
# setup experiment
self.num_epochs = self._trial.suggest_int("exp.num_epochs", 1, self.max_epochs)
Expand All @@ -99,20 +103,23 @@ def on_experiment_start(self, exp: "IExperiment"):
),
}
# setup model
hidden_size = self._trial.suggest_int("lstm.hidden_size", 32, 256, log=True)
num_layers = self._trial.suggest_int("lstm.num_layers", 1, 4)
bidirectional = self._trial.suggest_categorical("lstm.bidirectional", [True, False])
fc_dropout = self._trial.suggest_uniform("lstm.fc_dropout", 0.1, 0.9)
self.model = LSTM(
input_size=53, # PRIOR
hidden_size=self._trial.suggest_int("lstm.hidden_size", 32, 256, log=True),
num_layers=self._trial.suggest_int("lstm.num_layers", 1, 4),
hidden_size=hidden_size,
num_layers=num_layers,
batch_first=True,
bidirectional=self._trial.suggest_categorical(
"lstm.bidirectional", [True, False]
),
fc_dropout=self._trial.suggest_uniform("lstm.fc_dropout", 0.1, 0.9),
bidirectional=bidirectional,
fc_dropout=fc_dropout,
)
lr = self._trial.suggest_float("adam.lr", 1e-5, 1e-3, log=True)
self.criterion = nn.BCEWithLogitsLoss()
self.optimizer = optim.Adam(
self.model.parameters(),
lr=self._trial.suggest_float("adam.lr", 1e-5, 1e-3, log=True),
lr=lr,
)
# setup callbacks
self.callbacks = {
Expand All @@ -131,6 +138,17 @@ def on_experiment_start(self, exp: "IExperiment"):
minimize=False,
),
}
self.wandb_logger.config.update(
{
"num_epochs": self.num_epochs,
"batch_size": self.batch_size,
"hidden_size": hidden_size,
"num_layers": num_layers,
"bidirectional": bidirectional,
"fc_dropout": fc_dropout,
"lr": lr,
}
)

def run_dataset(self) -> None:
all_scores, all_targets = [], []
Expand Down Expand Up @@ -159,9 +177,7 @@ def run_dataset(self) -> None:
y_test = np.hstack(all_targets)
y_score = np.hstack(all_scores)
y_pred = (y_score > 0.5).astype(np.int32)
report = get_classification_report(
y_true=y_test, y_pred=y_pred, y_score=y_score, beta=0.5
)
report = get_classification_report(y_true=y_test, y_pred=y_pred, y_score=y_score, beta=0.5)
for stats_type in [0, 1, "macro", "weighted"]:
stats = report.loc[stats_type]
for key, value in stats.items():
Expand All @@ -174,13 +190,30 @@ def run_dataset(self) -> None:
"loss": total_loss,
}

def on_epoch_end(self, exp: "IExperiment") -> None:
super().on_epoch_end(self)
self.wandb_logger.log(
{
"train_score": self.epoch_metrics["train"]["score"],
"train_accuracy": self.epoch_metrics["train"]["accuracy"],
"train_loss": self.epoch_metrics["train"]["loss"],
"valid_score": self.epoch_metrics["valid"]["score"],
"valid_accuracy": self.epoch_metrics["valid"]["accuracy"],
"valid_loss": self.epoch_metrics["valid"]["loss"],
},
)

def on_experiment_end(self, exp: "IExperiment") -> None:
super().on_experiment_end(exp)
self._score = self.callbacks["early-stop"].best_score

wandb.summary["valid_score"] = self._score
self.wandb_logger.finish()

def _objective(self, trial) -> float:
self._trial = trial
self.run()

return self._score

def tune(self, n_trials: int):
Expand Down
49 changes: 42 additions & 7 deletions src/scripts/tune_ts_mlp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# pylint: disable-all
import argparse

from animus import EarlyStoppingCallback, IExperiment
Expand All @@ -16,6 +17,8 @@
from src.settings import LOGS_ROOT, UTCNOW
from src.ts import load_ABIDE1, TSQuantileTransformer

import wandb


class ResidualBlock(nn.Module):
def __init__(self, block):
Expand Down Expand Up @@ -105,6 +108,9 @@ def on_tune_start(self):
)

def on_experiment_start(self, exp: "IExperiment"):
# init wandb logger
self.wandb_logger: wandb.run = wandb.init(project="tune_mlp", name=f"{UTCNOW}-mlp")

super().on_experiment_start(exp)
# setup experiment
self.num_epochs = self._trial.suggest_int("exp.num_epochs", 1, self.max_epochs)
Expand All @@ -119,17 +125,22 @@ def on_experiment_start(self, exp: "IExperiment"):
),
}
# setup model
hidden_size = self._trial.suggest_int("mlp.hidden_size", 32, 256, log=True)
num_layers = self._trial.suggest_int("mlp.num_layers", 0, 4)
dropout = self._trial.suggest_uniform("mlp.dropout", 0.1, 0.9)
self.model = MLP(
input_size=53, # PRIOR
output_size=2, # PRIOR
hidden_size=self._trial.suggest_int("mlp.hidden_size", 32, 256, log=True),
num_layers=self._trial.suggest_int("mlp.num_layers", 0, 4),
dropout=self._trial.suggest_uniform("mlp.dropout", 0.1, 0.9),
hidden_size=hidden_size,
num_layers=num_layers,
dropout=dropout,
)

lr = self._trial.suggest_float("adam.lr", 1e-5, 1e-3, log=True)
self.criterion = nn.CrossEntropyLoss()
self.optimizer = optim.Adam(
self.model.parameters(),
lr=self._trial.suggest_float("adam.lr", 1e-5, 1e-3, log=True),
lr=lr,
)
# setup callbacks
self.callbacks = {
Expand All @@ -149,6 +160,17 @@ def on_experiment_start(self, exp: "IExperiment"):
),
}

self.wandb_logger.config.update(
{
"num_epochs": self.num_epochs,
"batch_size": self.batch_size,
"hidden_size": hidden_size,
"num_layers": num_layers,
"dropout": dropout,
"lr": lr,
}
)

def run_dataset(self) -> None:
all_scores, all_targets = [], []
total_loss = 0.0
Expand All @@ -173,9 +195,7 @@ def run_dataset(self) -> None:
y_test = np.hstack(all_targets)
y_score = np.vstack(all_scores)
y_pred = np.argmax(y_score, axis=-1).astype(np.int32)
report = get_classification_report(
y_true=y_test, y_pred=y_pred, y_score=y_score, beta=0.5
)
report = get_classification_report(y_true=y_test, y_pred=y_pred, y_score=y_score, beta=0.5)
for stats_type in [0, 1, "macro", "weighted"]:
stats = report.loc[stats_type]
for key, value in stats.items():
Expand All @@ -187,13 +207,28 @@ def run_dataset(self) -> None:
"loss": total_loss,
}

def on_epoch_end(self, exp: "IExperiment") -> None:
super().on_epoch_end(self)
self.wandb_logger.log(
{
"train_score": self.epoch_metrics["train"]["score"],
"train_loss": self.epoch_metrics["train"]["loss"],
"valid_score": self.epoch_metrics["valid"]["score"],
"valid_loss": self.epoch_metrics["valid"]["loss"],
},
)

def on_experiment_end(self, exp: "IExperiment") -> None:
super().on_experiment_end(exp)
self._score = self.callbacks["early-stop"].best_score

wandb.summary["valid_score"] = self._score
self.wandb_logger.finish()

def _objective(self, trial) -> float:
self._trial = trial
self.run()

return self._score

def tune(self, n_trials: int):
Expand Down
Loading