def max_pool_2d(x, kernelshape, pad=(0, 0), stride=(1, 1)):
devtype = cgt.get_config()["default_device"].devtype
kernel_h, kernel_w = kernelshape
pad_h, pad_w = pad
stride_h, stride_w = stride
info = PoolInfo(kernel_h, kernel_w, pad_h, pad_w, stride_h, stride_w)
if devtype == "gpu":
return core.Result(cudnn_ops.CudnnPoolForward(info), [x])
else:
return core.Result(MaxPool(info), [x])[0]
def max_pool_2d(x, kernelshape, pad = (0,0), stride=(1,1)):
devtype = cgt.get_config()["default_device"].devtype
kernel_h, kernel_w = kernelshape
pad_h, pad_w = pad
stride_h, stride_w = stride
info = PoolInfo(kernel_h, kernel_w, pad_h, pad_w, stride_h, stride_w)
if devtype == "gpu":
return core.Result(cudnn_ops.CudnnPoolForward(info), [x])
else:
return core.Result(MaxPool(info), [x])[0]
def conv2d(x_BKRC, f_LKrc, kernelshape, pad=(0,0), stride=(1,1)):
devtype = cgt.get_config()["default_device"].devtype
if devtype == "gpu":
return cudnn_ops.CudnnConvForward(pad[0],pad[1],stride[0],stride[1])
else:
assert devtype == "cpu"
col_BmnZ = im2col(x_BKRC, kernelshape, pad, stride)
L,K,r,c = f_LKrc.shape
f_LZ = f_LKrc.reshape([L, K*r*c])
B,m,n,Z = col_BmnZ.shape
col_Bmn_Z = col_BmnZ.reshape([B*m*n, Z])
col_Bmn_L = core.Result(core.Mul22(False,True), [col_Bmn_Z, f_LZ])
return col_Bmn_L.reshape([B,m,n,L]).transpose([0,3,1,2])
def runTest(self):
if cgt.get_config()["backend"] != "python":
cgt.utils.warn("Skipping test -- only works for backend=python")
return
x = cgt.scalar()
with cgt.debug_context() as dbg:
cgt.assert_(cgt.equal(x, 1), "yoyoyo")
cgt.dbg_call(myfunc, x)
print "dbg", dbg.nodes
# cgt.assert_(cgt.equal(x, 2))
f = cgt.make_function([x], [x], dbg=dbg)
f(1)
with self.assertRaises(AssertionError):
f(2)
def runTest(self):
if cgt.get_config()["backend"] != "python":
cgt.utils.warn("Skipping test -- only works for backend=python")
return
x = cgt.scalar()
with cgt.debug_context() as dbg:
cgt.assert_(cgt.equal(x, 1),"yoyoyo")
cgt.dbg_call(myfunc, x)
print "dbg",dbg.nodes
# cgt.assert_(cgt.equal(x, 2))
f = cgt.make_function([x],[x],dbg=dbg)
f(1)
with self.assertRaises(AssertionError):
f(2)
def conv2d(x_BKRC, f_LKrc, kernelshape, pad=(0, 0), stride=(1, 1)):
devtype = cgt.get_config()["default_device"].devtype
L, K, r, c = f_LKrc.shape
if devtype == "gpu":
b_1K11 = cgt.zeros((1, L, 1, 1), cgt.floatX)
return core.Result(
cudnn_ops.CudnnConvForward(pad[0], pad[1], stride[0], stride[1]),
[x_BKRC, f_LKrc, b_1K11])
else:
assert devtype == "cpu"
col_BmnZ = im2col(x_BKRC, kernelshape, pad, stride)
f_LZ = f_LKrc.reshape([L, K * r * c])
B, m, n, Z = col_BmnZ.shape
col_Bmn_Z = col_BmnZ.reshape([B * m * n, Z])
col_Bmn_L = core.Result(core.Mul22(False, True), [col_Bmn_Z, f_LZ])
return col_Bmn_L.reshape([B, m, n, L]).transpose([0, 3, 1, 2])
def create_interpreter(inputs, outputs, eg, node2memloc):
assert isinstance(eg, ExecutionGraph)
input_types = [input.typ for input in inputs] #pylint: disable=W0622
output_locs = [node2memloc[node] for node in outputs]
config = cgt.get_config()
backend = config["backend"]
parallel = config["parallel"]
if backend == "python":
if parallel:
raise NotImplementedError("For parallel=True, set backend=native")
# return ParallelInterpreter(eg, output_locs, input_types)
else:
return SequentialInterpreter(eg, output_locs, input_types)
elif backend == "native":
if parallel:
return cgt.cycgt.CppInterpreterWrapper(eg, input_types, output_locs, config["num_threads"])
else:
return cgt.cycgt.CppInterpreterWrapper(eg, input_types, output_locs, 0)
else:
raise NotImplementedError("invalid backend %s"%backend)
import rnn
from utils.opt import *
from utils.debug import safe_io
from utils.utilities import NONE
# import traceback
# def _numpy_err_callback(type, flag):
# print type, flag
# traceback.print_stack()
# raise FloatingPointError('refer to _numpy_err_callback for more details')
# np.seterr(divide='call', over='warn', invalid='call', under='warn')
# np.seterrcall(_numpy_err_callback)
# np.set_printoptions(precision=4, suppress=True)
cgt.check_source() # this line will fail if CGT in use is not TZ2016's fork
print cgt.get_config(True)
def init(args):
ws = {}
ws['config'] = copy.deepcopy(args)
_is_sto = any(_n != 0 for _n in args['num_sto'])
_is_rec = any(_n != 0 for _n in args['num_mems'])
assert not (_is_rec and _is_sto), "Stochastic recurrent units not supported"
net_type = []
if _is_sto: net_type.append('snn')
else: net_type.append('dnn')
if _is_rec: net_type.append('rnn')
else: net_type.append('fnn')
ws['type'] = net_type
# TODO: add in the dbg_out
import sfnn
import rnn
from utils.opt import *
from utils.debug import safe_io
from utils.utilities import NONE
# import traceback
# def _numpy_err_callback(type, flag):
# print type, flag
# traceback.print_stack()
# raise FloatingPointError('refer to _numpy_err_callback for more details')
# np.seterr(divide='call', over='warn', invalid='call', under='warn')
# np.seterrcall(_numpy_err_callback)
# np.set_printoptions(precision=4, suppress=True)
cgt.check_source() # this line will fail if CGT in use is not TZ2016's fork
print cgt.get_config(True)
def init(args):
ws = {}
ws['config'] = copy.deepcopy(args)
_is_sto = any(_n != 0 for _n in args['num_sto'])
_is_rec = any(_n != 0 for _n in args['num_mems'])
assert not (_is_rec
and _is_sto), "Stochastic recurrent units not supported"
net_type = []
if _is_sto: net_type.append('snn')
else: net_type.append('dnn')
if _is_rec: net_type.append('rnn')
else: net_type.append('fnn')
ws['type'] = net_type
