File size: 3,644 Bytes
eb90a8a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 | """
Module defining some utility functions used with ML applications
"""
import os
import sys
import numpy as np
import tensorflow as tf
import random as python_random
def resetSeeds():
"""
Reset system seeds for random generators
"""
np.random.seed(123)
python_random.seed(123)
tf.random.set_seed(1234)
def setupCheckpoint(checkpoint_dir):
"""
Set up writing down checkpoints:
- Make checkpoint directory is it does not exist
- Find latest checkpoint
Parameter:
- checkpoint_dir -- Directory for writing down checkpoints
Returns:
"""
if not os.path.exists(checkpoint_dir):
os.makedirs(checkpoint_dir)
latest_checkpoint = None
else:
checkpoints = [checkpoint_dir + "/" + name for
name in os.listdir(checkpoint_dir)]
latest_checkpoint = max(checkpoints, key=os.path.getctime) \
if checkpoints else None
return latest_checkpoint
def getCheckpoint(checkpoint_dir, checkpoint = None):
"""
Get either the specified checkpoint or the latest one
Parameters:
- checkpoint_dir -- Directory with checkpoints
- checkpoint -- Checkpoint file name
Returns: full path to checkpoint file
"""
if not checkpoint_dir:
sys.exit("Directory with DNN models is not defined")
if not os.path.exists(checkpoint_dir):
sys.exit(f"Directory {checkpoint_dir} with DNN models does not exist")
if checkpoint:
if os.path.exists(f"{checkpoint_dir}/checkpoint"):
return ckpt_file
else:
sys.exit(f"Directory {checkpoint_dir} does not have checkpoint {checkpoint}")
checkpoints = [checkpoint_dir + "/" + name for
name in os.listdir(checkpoint_dir)]
if checkpoints:
return max(checkpoints, key=os.path.getctime)
else:
sys.exit(f"Directory {checkpoint_dir} with DNN models is empty")
def makeCkptCallback(checkpoint_dir, monitor = "val_accuracy"):
"""
- Create callback function for writing down checkpoints
Parameter:
- checkpoint_dir -- Directory for writing down checkpoints
- monitor -- value to monitor
Returns:
- Callback function created
"""
checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(
filepath = checkpoint_dir + "/ckpt-{epoch}",
monitor = monitor,
verbose = 0, save_best_only = True,
save_weights_only = False,
mode = "auto", save_freq= "epoch")
return checkpoint_callback
def memoryUsage(point):
"""
Print memory usage
Parameter:
- point -- identifier of print: string or number
"""
from resource import getrusage, RUSAGE_SELF
_resource_usage = getrusage(RUSAGE_SELF)
print(f"{point}: MEMORY = {_resource_usage.ru_maxrss}")
def makeFilePath(*parts, m = None):
"""
Make path to file and check its existance
Check if the path exists and exit if it is not found
The existance of the path is checked only if
the description of file is defined
Parameters:
- m -- description of the path to pringt in the message
if the path does not exist
- parts -- parts to concatenate to get file path
"""
_p = "/".join(parts).replace("//", "/")
if m and not os.path.exists(_p):
sys.exit(f"{m} {_p} is not found")
return _p
'''
def accessModules(*modules):
"""
Update sys.path to access required modules
Parameters:
- modules -- modules to access
"""
main_dir = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
'''
|