def set_configuration(self, cfg_in):
import torch
from bioharn import clf_predict
cfg = self.get_configuration()
_vital_config_update(cfg, cfg_in)
for key in self._kwiver_config.keys():
self._kwiver_config[key] = str(cfg.get_value(key))
if self._kwiver_config['batch_size'] == "auto":
self._kwiver_config['batch_size'] = 2
if torch.cuda.is_available():
gpu_mem = 0
if len(self._kwiver_config['xpu']) == 1 and \
self._kwiver_config['xpu'] != 0:
gpu_id = int(self._kwiver_config['xpu'])
gpu_mem = torch.cuda.get_device_properties(
gpu_id).total_memory
else:
self._gpu_count = torch.cuda.device_count()
for i in range(self._gpu_count):
single_gpu_mem = torch.cuda.get_device_properties(
i).total_memory
if gpu_mem == 0:
gpu_mem = single_gpu_mem
else:
gpu_mem = min(gpu_mem, single_gpu_mem)
if gpu_mem > 9e9:
self._kwiver_config['batch_size'] = 4
elif gpu_mem >= 7e9:
self._kwiver_config['batch_size'] = 3
pred_config = clf_predict.ClfPredictConfig()
pred_config['batch_size'] = self._kwiver_config['batch_size']
pred_config['deployed'] = self._kwiver_config['deployed']
pred_config['xpu'] = self._kwiver_config['xpu']
pred_config['input_dims'] = 'native'
# (256, 256)
self.predictor = clf_predict.ClfPredictor(pred_config)
self.predictor._ensure_model()
self._area_pivot = int(self._kwiver_config['area_pivot'])
self._area_lower_bound = int(self._kwiver_config['area_lower_bound'])
self._area_upper_bound = int(self._kwiver_config['area_upper_bound'])
self._border_exclude = int(self._kwiver_config['border_exclude'])
self._average_prior = strtobool(self._kwiver_config['average_prior'])
if self._area_pivot < 0:
self._area_upper_bound = -self._area_pivot
elif self._area_pivot > 0:
self._area_lower_bound = self._area_pivot
return True
def set_configuration(self, cfg_in):
import torch
from bioharn import clf_predict
cfg = self.get_configuration()
# HACK: merge config doesn't support dictionary input
_vital_config_update(cfg, cfg_in)
for key in self._kwiver_config.keys():
self._kwiver_config[key] = str(cfg.get_value(key))
if self._kwiver_config['batch_size'] == "auto":
self._kwiver_config['batch_size'] = 2
if torch.cuda.is_available():
gpu_mem = 0
if len(self._kwiver_config['xpu']) == 1 and \
self._kwiver_config['xpu'] != 0:
gpu_id = int(self._kwiver_config['xpu'])
gpu_mem = torch.cuda.get_device_properties(
gpu_id).total_memory
else:
self._gpu_count = torch.cuda.device_count()
for i in range(self._gpu_count):
single_gpu_mem = torch.cuda.get_device_properties(
i).total_memory
if gpu_mem == 0:
gpu_mem = single_gpu_mem
else:
gpu_mem = min(gpu_mem, single_gpu_mem)
if gpu_mem > 9e9:
self._kwiver_config['batch_size'] = 4
elif gpu_mem >= 7e9:
self._kwiver_config['batch_size'] = 3
pred_config = clf_predict.ClfPredictConfig()
pred_config['batch_size'] = self._kwiver_config['batch_size']
pred_config['deployed'] = self._kwiver_config['deployed']
if torch.cuda.is_available():
pred_config['xpu'] = self._kwiver_config['xpu']
else:
pred_config['xpu'] = "cpu"
pred_config['input_dims'] = 'native'
# (256, 256)
self.predictor = clf_predict.ClfPredictor(pred_config)
self.predictor._ensure_model()
return True
def set_configuration(self, cfg_in):
import torch
from bioharn import clf_predict
cfg = self.get_configuration()
_vital_config_update(cfg, cfg_in)
for key in self._kwiver_config.keys():
self._kwiver_config[key] = str(cfg.get_value(key))
if self._kwiver_config['batch_size'] == "auto":
self._kwiver_config['batch_size'] = 2
if torch.cuda.is_available():
gpu_mem = 0
if len(self._kwiver_config['xpu']) == 1 and \
self._kwiver_config['xpu'] != 0:
gpu_id = int(self._kwiver_config['xpu'])
gpu_mem = torch.cuda.get_device_properties(
gpu_id).total_memory
else:
self._gpu_count = torch.cuda.device_count()
for i in range(self._gpu_count):
single_gpu_mem = torch.cuda.get_device_properties(
i).total_memory
if gpu_mem == 0:
gpu_mem = single_gpu_mem
else:
gpu_mem = min(gpu_mem, single_gpu_mem)
if gpu_mem > 9e9:
self._kwiver_config['batch_size'] = 4
elif gpu_mem >= 7e9:
self._kwiver_config['batch_size'] = 3
pred_config = clf_predict.ClfPredictConfig()
pred_config['batch_size'] = self._kwiver_config['batch_size']
pred_config['deployed'] = self._kwiver_config['deployed']
pred_config['xpu'] = self._kwiver_config['xpu']
pred_config['input_dims'] = 'native' # (256, 256)
self.predictor = clf_predict.ClfPredictor(pred_config)
self.predictor._ensure_model()
self._area_pivot = int(self._kwiver_config['area_pivot'])
self._area_lower_bound = int(self._kwiver_config['area_lower_bound'])
self._area_upper_bound = int(self._kwiver_config['area_upper_bound'])
self._border_exclude = int(self._kwiver_config['border_exclude'])
self._average_prior = strtobool(self._kwiver_config['average_prior'])
if self._area_pivot < 0:
self._area_upper_bound = -self._area_pivot
elif self._area_pivot > 0:
self._area_lower_bound = self._area_pivot
# Load scale based on type file if enabled
self._target_type_scales = dict()
if self._kwiver_config['scale_type_file']:
fin = open(self._kwiver_config['scale_type_file'], 'r')
for line in fin.readlines():
line = line.rstrip()
parsed_line = line.split()
if len(parsed_line) < 1:
continue
target_area = float(parsed_line[-1])
type_str = str(' '.join(parsed_line[:-1]))
self._target_type_scales[type_str] = target_area
return True