def model_files_by_timestamp(dirpath):
"""
List all available model files by timestamp, most recent first
"""
models = {}
model_files = read_dir(dirpath)
for f in model_files:
filepath = dirpath + '/' + f
ts = os.path.getmtime(filepath)
models[ts] = filepath
return sorted(
models.items(), key=lambda x: x[0], reverse=True
) # "reverse", because we want the highest timestamps (most recent) first
def model_files_by_F1(dirpath):
"""
List all available model files by loss
"""
models = {}
model_dirs = read_dir(dirpath)
for f in model_dirs:
fv = re.sub(r'^F', '', f)
fv = re.sub(r'_E\d+_B\d+$', '', fv)
F_val = float(fv)
modelpath = dirpath + '/' + f
models[modelpath] = F_val
return sorted(models.items(), key=lambda x: x[1], reverse=True)
def model_files_by_loss(dirpath):
"""
List all available model files by loss
"""
models = {}
model_dirs = read_dir(dirpath)
for f in model_dirs:
lv = re.sub(r'^F', '', f)
lv = re.sub(r'_E\d+_B\d+$', '', lv)
loss_val = float(lv)
modelpath = dirpath + '/' + f
models[modelpath] = loss_val
return sorted(models.items(), key=lambda x: x[1])
def model_files_by_loss(self):
"""
List all available model files by loss
"""
models = {}
model_files = read_dir(self.get('model_dir'))
for f in model_files:
if re.search('^loss_', f):
lv = re.sub(r'^loss_', '', f)
lv = re.sub(r'\.json\.gz$', '', lv)
loss_val = float(lv)
models[loss_val] = self.get('model_dir') + '/' + f
return sorted(models.items(), key=lambda x: x[0])
def model_files_by_timestamp(self):
"""
List all available model files by timestamp, most recent first
"""
models = {}
model_files = read_dir(self.get('model_dir'))
for f in model_files:
if re.search('^loss_', f):
filepath = self.get('model_dir') + '/' + f
ts = os.path.getmtime(filepath)
models[ts] = filepath
return sorted(
models.items(), key=lambda x: x[0], reverse=True
) # "reverse", because we want the highest timestamps (most recent) first
def default_read_data_dir(input_dir):
"""
Default function for reading data from a directory into a Dataset
Parameters
----------
input_dir : str
Directory where each file will be read
Returns
-------
array of DataFrame (4 of them: x_train, x_test, y_train, y_test)
"""
files = read_dir(input_dir, {'fullpath': True})
return default_read_data_files(files)
def model_files_by_pcc(dirpath):
"""
List all available model files by PCC
"""
models = {}
model_dirs = read_dir(dirpath)
for f in model_dirs:
if re.search(r'^PCC', f):
lv = re.sub(r'^PCC', '', f)
lv = re.sub(r'_E\d+_B\d+$', '', lv)
try:
loss_val = float(lv)
modelpath = dirpath + '/' + f
models[modelpath] = loss_val
except:
pass
return sorted(models.items(), key=lambda x: x[1], reverse=True)