here is two module sequential
contrastive_seq_a = orch.nn.Sequential(
#torch.nn.Linear(2 * 46 * 296, 500),
torch.nn.Linear(2 * 26 * 296, 500),
torch.nn.BatchNorm1d(500),
torch.nn.ReLU(),
torch.nn.Dropout(self.config.dropout),
torch.nn.Linear(500, 10),
#torch.nn.BatchNorm1d(10),
#torch.nn.ReLU(),
#torch.nn.Dropout(self.config.dropout),
torch.nn.Linear(10, 2),
# torch.nn.Sigmoid()
)
contrastive_seq_b = torch.nn.Sequential(
#torch.nn.Linear(2 * 46 * 296, 500),
torch.nn.Linear(2 * 26 * 296, 500),
torch.nn.BatchNorm1d(500),
torch.nn.ReLU(),
torch.nn.Dropout(self.config.dropout),
torch.nn.Linear(500, 10),
#torch.nn.BatchNorm1d(10),
#torch.nn.ReLU(),
torch.nn.Dropout(self.config.dropout),
torch.nn.Linear(10, 2),
# torch.nn.Sigmoid()
)
when i use contrastive_seq_b ,i got err after several iterations
err is “RuntimeError: cuda runtime error (59) : device-side assert triggered at /data/TENCENT.COM/nealchai/download/pytorch/aten/src/THC/generic/THCTensorScatterGather.cu:75”
the train data is no problem。
I print the network parameters，it becames nan when the err occurs。
my loss function is ContrastiveLoss
class ContrastiveLoss(torch.nn.Module):
"""
Contrastive loss function.
Based on:
"""
def __init__(self, margin=1.0):
super(ContrastiveLoss, self).__init__()
self.margin = margin
def check_type_forward(self, in_types):
assert len(in_types) == 3
x0_type, x1_type, y_type = in_types
# print("dim:{},{}".format(x0_type.dim(), x0_type.shape))
assert x0_type.size() == x1_type.shape
assert x1_type.size()[0] == y_type.shape[0]
assert x1_type.size()[0] > 0
assert x0_type.dim() == 2
assert x1_type.dim() == 2
assert y_type.dim() == 1
def forward(self, x0, x1, y):
self.check_type_forward((x0, x1, y))
# euclidian distance
diff = x0 - x1
dist_sq = torch.sum(torch.pow(diff, 2), 1)
dist = torch.sqrt(dist_sq)
mdist = self.margin - dist
dist = torch.clamp(mdist, min=0.0)
loss = y * dist_sq + (1 - y) * torch.pow(dist, 2)
loss = torch.sum(loss) / 2.0 / x0.size()[0]
return loss