def main(cfg: DictConfig):
session = Session(
repo=(Path.home() / "projects/rlbook/experiments/outputs/bandit").as_posix(),
experiment=cfg.bandit["_target_"].split(".")[-1],
)
testbed = instantiate(cfg.testbed)
bandit = instantiate(cfg.bandit, Q_init=call(cfg.Q_init, testbed))
local_logger.info(f"Running bandit: {cfg.run}")
local_logger.debug(f"Testbed expected values: {testbed.expected_values}")
local_logger.debug(f"bandit config: {cfg['bandit']}")
local_logger.debug(f"run config: {cfg['run']}")
session.set_params(OmegaConf.to_container(cfg.run), "experiment")
bandit.run(testbed, **OmegaConf.to_container(cfg.run))
df_ar = bandit.output_df()
df_ar = optimal_action(df_ar)
local_logger.debug(f"\n{df_ar[['run', 'step', 'action', 'optimal_action', 'reward']].head(15)}")
bandit_type = cfg.bandit._target_.split(".")[-1]
Q_init = cfg.Q_init._target_.split(".")[-1]
task_name = f"{bandit_type} - " + ", ".join(
[
f"{k}: {OmegaConf.select(cfg, v).split('.')[-1]}"
if isinstance(OmegaConf.select(cfg, v), str)
else f"{k}: {OmegaConf.select(cfg, v)}"
for k, v in cfg.task_labels.items()
]
)
local_logger.debug(f"{task_name}")
hp_testbed = OmegaConf.to_container(cfg.testbed)
hp = OmegaConf.to_container(cfg.bandit)
hp["Q_init"] = cfg.Q_init._target_
hp["p_drift"] = hp_testbed["p_drift"]
session.set_params(hp, "hyperparameters")
# for i in range(min(3, cfg.run.n_runs)):
# fig = steps_violin_plotter(df_ar, testbed, run=i)
# writer.add_figure(f"run{i}", fig, global_step=cfg.run.steps)
final_avg_reward = write_scalars(df_ar, session, "reward", "average_reward", hp)
final_optimal_action = write_scalars(
df_ar, session, "optimal_action_percent", "optimal_action_percent", hp
)
final_metrics = {
"average_reward": final_avg_reward,
"optimal_action_percent": final_optimal_action,
}
session.set_params(final_metrics, "final_metrics")
local_logger.debug(f"final_metrics: {final_metrics}")
from aim import Session
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
# Create Aim Session
aim_sess = Session()
# Device configuration
device = torch.device('cpu')
# Hyper parameters
num_epochs = 5
num_classes = 10
batch_size = 50
learning_rate = 0.01
# aim - Track hyper parameters
aim_sess.set_params(
{
'num_epochs': num_epochs,
'num_classes': num_classes,
'batch_size': batch_size,
'learning_rate': learning_rate,
},
name='hparams')
# MNIST dataset
train_dataset = torchvision.datasets.MNIST(root='./data/',
model = Sequential()
# Conv block
model.add(
Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=input_shape))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
# Dense block
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))
model.compile(loss=keras.losses.categorical_crossentropy,
optimizer=keras.optimizers.Adadelta(),
metrics=['accuracy'])
model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(x_test, y_test),
callbacks=[
AimCallback(session=Session(experiment='test_keras_cb')),
])
# score = model.evaluate(x_test, y_test, verbose=0)
from aim.tensorflow import AimCallback
from aim import Session
import tensorflow as tf
mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(28, 28)),
tf.keras.layers.Dense(128, activation='relu'),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(10)
])
loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)
model.compile(optimizer='adam', loss=loss_fn, metrics=['accuracy'])
model.fit(
x_train,
y_train,
epochs=5,
callbacks=[AimCallback(session=Session(experiment='test_tf_1_keras_cb'))])
from aim import Session
sess = Session()
sess.set_params({
'foo': 'bar',
})
for i in range(10):
sess.track(i, name='val')
from aim import Session
sess1 = Session(experiment='line')
sess2 = Session(experiment='linex2')
sess1.set_params({
'k': '1',
})
sess2.set_params({
'k': '2',
})
for i in range(10):
sess1.track(i, name='val')
sess2.track(i*2, name='val')