def _worker(i, module, input, kwargs, device=None):
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
output = module(input, **kwargs)
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, target, kwargs, device=None):
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
if not isinstance(input, (list, tuple)):
input = (input, )
output = module(*(input + target), **kwargs)
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, target, kwargs, device=None):
if torch_ver != "0.3":
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
output = module(input, target)
# output = module(*(input + target), **kwargs)# modified at Issue 54
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, kwargs, device=None):
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
# this also avoids accidental slicing of `input` if it is a Tensor
if not isinstance(input, (list, tuple)):
input = (input, )
output = getattr(module, func_name)(*input, **kwargs)
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, target, kwargs, device=None):
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
if not isinstance(input, (list, tuple)):
input = (input, )
if not isinstance(target, (list, tuple)):
target = (target, )
output = module(*input, *target, **kwargs)
with lock:
results[i] = output
except Exception:
with lock:
results[i] = ExceptionWrapper(
where="in replica {} on device {}".format(i, device))
def _worker(i, module, input, target, kwargs, device=None):
if torch_ver != "0.3":
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
# output = module(*(input + target), **kwargs)
if type(input) == tuple:
input = input[0]
# print('module: {}\ninput: {}\ntarget:{}'.format(module, input, target))
output = module(input, target)
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, target, kwargs, device=None):
#import pdb;pdb.set_trace()
if torch_ver != "0.3":
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
if not isinstance(input, tuple):
input = (input, )
if not isinstance(target, tuple):
target = (target, )
with torch.cuda.device(device):
output = module(*(input + target), **kwargs)
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, target, kwargs, device=None):
if torch_ver != "0.3":
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
# this also avoids accidental slicing of `input` if it is a Tensor
if not isinstance(input, (list, tuple)):
input = (input,)
if not isinstance(target, (list, tuple)):
target = (target,)
output = module(*(input + target)) # **kwargs) removed by Turing_Lee
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, kwargs, device=None):
if torch_ver != "0.3":
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
# this also avoids accidental slicing of `input` if it is a Tensor
if not isinstance(input, (list, tuple)):
input = (input,)
# FTWS: Notice that in Decoder and parallel loss the input has already been tuple, so
# we change codes here.
output = module(input, **kwargs)
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, kwargs, device=None):
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
# this also avoids accidental slicing of `input` if it is a Tensor
if not isinstance(input, (list, tuple)):
input = (input,)
if forward:
output = module(*input, **kwargs)
else:
output = module.sample(*input, **kwargs)
with lock:
results[i] = output
except Exception:
with lock:
results[i] = ExceptionWrapper(
where="in replica {} on device {}".format(i, device))
def _worker(i, module, input, target, kwargs, device=None):
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
# PyTorch's native implementation convert to tuple to avoid further slicing
# Just extract the tensor out of the tuple to compute loss
if isinstance(input, (list, tuple)):
input = input[0]
if isinstance(target, (list, tuple)):
target = target[0]
assert target.device == input.device
if module.device != input.device:
module = module.to(input.device)
output = module(input, target, **kwargs)
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
def _worker(i, module, input, target, kwargs, device=None):
if torch_ver != "0.3":
torch.set_grad_enabled(grad_enabled)
if device is None:
device = get_a_var(input).get_device()
try:
with torch.cuda.device(device):
# this also avoids accidental slicing of `input` if it is a Tensor
if not isinstance(input, (list, tuple)):
input = (input, )
if type(input) != type(target):
if isinstance(target, tuple):
input = tuple(input)
elif isinstance(target, list):
input = list(input)
else:
raise Exception("Types problem")
output = module(*(input + target), **kwargs)
with lock:
results[i] = output
except Exception as e:
with lock:
results[i] = e
