def train(self, solverFilename, logFile, weights=None):
# hackaround: this backend doesn't log to one file but to a db in a subdirectory
import caffe
caffe.setup_teeing(logFile)
from pymill.CNN import MillSolver as ms
self.solver = ms.MillSolver(solver_def=solverFilename, weights=weights, gpus=[int(x) for x in self._gpus.split(',')])
self.solver.run_solver()
def resume(self, solverFilename, solverstateFilename, logFile):
# hackaround: this backend doesn't log to one file but to a db in a subdirectory
import caffe
caffe.setup_teeing(logFile)
log_dir = os.path.abspath(logFile)
from pymill.CNN import MillSolver as ms
self.solver = ms.MillSolver(solver_def=solverFilename, solver_state=solverstateFilename, gpus=[int(x) for x in self._gpus.split(',')])
self.solver.run_solver()
def train(self, solverFilename, logFile, weights=None):
# hackaround: this backend doesn't log to one file but to a db in a subdirectory
import caffe
caffe.setup_teeing(logFile)
from pymill.CNN import MillSolver as ms
self.solver = ms.MillSolver(
solver_def=solverFilename,
weights=weights,
gpus=[int(x) for x in self._gpus.split(',')])
self.solver.run_solver()
def resume(self, solverFilename, solverstateFilename, logFile):
# hackaround: this backend doesn't log to one file but to a db in a subdirectory
import caffe
caffe.setup_teeing(logFile)
log_dir = os.path.abspath(logFile)
from pymill.CNN import MillSolver as ms
self.solver = ms.MillSolver(
solver_def=solverFilename,
solver_state=solverstateFilename,
gpus=[int(x) for x in self._gpus.split(',')])
self.solver.run_solver()
