import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
import argparse
from seisnn.core import Feature
from seisnn.utils import get_config
from seisnn.io import read_dataset
from seisnn.plot import plot_loss
from seisnn.example_proto import batch_iterator
ap = argparse.ArgumentParser()
ap.add_argument('-m', '--model', required=False, help='model', type=str)
args = ap.parse_args()
config = get_config()
SAVE_MODEL_PATH = os.path.join(config['MODELS_ROOT'], args.model)
SAVE_HISTORY_PATH = os.path.join(SAVE_MODEL_PATH, 'history')
SAVE_PNG_PATH = os.path.join(SAVE_MODEL_PATH, 'png')
loss_log = os.path.join(SAVE_MODEL_PATH, f'{args.model}.log')
plot_loss(loss_log, SAVE_MODEL_PATH)
dataset = read_dataset(SAVE_HISTORY_PATH)
for batch in dataset.batch(2):
for example in batch_iterator(batch):
feature = Feature(example)
feature.plot(title=feature.id, save_dir=SAVE_PNG_PATH)
print(feature.id)
'--pre_train',
required=True,
help='pre-train model',
type=str)
ap.add_argument('-m', '--model', required=True, help='save model', type=str)
args = ap.parse_args()
config = get_config()
SAVE_MODEL_PATH = os.path.join(config['MODELS_ROOT'], args.model)
make_dirs(SAVE_MODEL_PATH)
SAVE_HISTORY_PATH = os.path.join(SAVE_MODEL_PATH, "history")
make_dirs(SAVE_HISTORY_PATH)
dataset_dir = os.path.join(config['DATASET_ROOT'], args.dataset)
dataset = read_dataset(dataset_dir).skip(1000)
val = next(iter(dataset.batch(1)))
val_trace = val['trace'][:, :, :, 0, tf.newaxis]
val_pdf = val['pdf'][:, :, :, 0, tf.newaxis]
ckpt = tf.train.Checkpoint(model=model, optimizer=optimizer)
ckpt_manager = tf.train.CheckpointManager(ckpt,
SAVE_MODEL_PATH,
max_to_keep=100)
if args.pre_train:
PRE_TRAIN_PATH = os.path.join(config['MODELS_ROOT'], args.pre_train)
for file in glob.glob(os.path.join(PRE_TRAIN_PATH, 'ckpt*')):
shutil.copy2(file, SAVE_MODEL_PATH)
required=True,
help='output dataset',
type=str)
ap.add_argument('-m', '--model', required=True, help='model', type=str)
args = ap.parse_args()
config = get_config()
MODEL_PATH = os.path.join(config['MODELS_ROOT'], args.model)
make_dirs(MODEL_PATH)
OUTPUT_DATASET = os.path.join(config['DATASET_ROOT'], args.output)
make_dirs(OUTPUT_DATASET)
INPUT_DATASET = os.path.join(config['DATASET_ROOT'], args.input)
dataset = read_dataset(INPUT_DATASET)
ckpt = tf.train.Checkpoint(model=model, optimizer=optimizer)
ckpt_manager = tf.train.CheckpointManager(ckpt, MODEL_PATH, max_to_keep=100)
if ckpt_manager.latest_checkpoint:
ckpt.restore(ckpt_manager.latest_checkpoint)
last_epoch = len(ckpt_manager.checkpoints)
print(f'Latest checkpoint epoch {last_epoch} restored!!')
n = 0
for batch in dataset.take(1000).batch(512).prefetch(2):
pdf = model.predict(batch['trace'])
batch['pdf'] = tf.concat([batch['pdf'], pdf], axis=3)
phase = batch['phase'].to_list()
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
import argparse
from seisnn.utils import get_config
from seisnn.io import read_dataset
from seisnn.core import Feature
from seisnn.example_proto import batch_iterator
ap = argparse.ArgumentParser()
ap.add_argument('-d', '--dataset', required=False, help='dataset', type=str)
args = ap.parse_args()
config = get_config()
dataset_dir = os.path.join(config['DATASET_ROOT'], args.dataset)
dataset = read_dataset(dataset_dir)
for batch in dataset.shuffle(1000).batch(2):
for example in batch_iterator(batch):
feature = Feature(example)
feature.plot(enlarge=True, snr=True)
ap = argparse.ArgumentParser()
ap.add_argument('-d', '--dataset', required=True, help='dataset', type=str)
ap.add_argument('-m', '--model', required=True, help='save model', type=str)
args = ap.parse_args()
config = get_config()
SAVE_MODEL_PATH = os.path.join(config['MODELS_ROOT'], args.model)
shutil.rmtree(SAVE_MODEL_PATH, ignore_errors=True)
make_dirs(SAVE_MODEL_PATH)
SAVE_HISTORY_PATH = os.path.join(SAVE_MODEL_PATH, "history")
make_dirs(SAVE_HISTORY_PATH)
dataset_dir = os.path.join(config['DATASET_ROOT'], args.dataset)
dataset = read_dataset(dataset_dir).shuffle(10000).take(1000)
val = next(iter(dataset.batch(1)))
val_trace = val['trace']
val_pdf = val['pdf']
ckpt = tf.train.Checkpoint(model=model, optimizer=optimizer)
ckpt_manager = tf.train.CheckpointManager(ckpt,
SAVE_MODEL_PATH,
max_to_keep=100)
EPOCHS = 1
for epoch in range(EPOCHS):
n = 0
loss_buffer = []
for train in dataset.prefetch(100).batch(1):
import pandas as pd
from obspy import UTCDateTime
from seisnn.core import Feature
from seisnn.utils import get_config
from seisnn.io import read_dataset
from seisnn.qc import signal_to_noise_ratio
from seisnn.plot import plot_snr_distribution
ap = argparse.ArgumentParser()
ap.add_argument('-d', '--dataset', required=False, help='dataset', type=str)
args = ap.parse_args()
config = get_config()
dataset_dir = os.path.join(config['DATASET_ROOT'], args.dataset)
dataset = read_dataset(dataset_dir).shuffle(100000).prefetch(10)
pick_snr = []
n = 0
for example in dataset:
feature = Feature(example)
picks = pd.DataFrame.from_dict({
'pick_time': feature.pick_time,
'pick_phase': feature.pick_phase,
'pick_set': feature.pick_set
})
picks = picks.loc[picks['pick_set'] == "manual"]
for i, p in picks.iterrows():
pick_time = UTCDateTime(p['pick_time']) - UTCDateTime(
feature.starttime)