私は非常に基本的な MLP ネットを持っています:
def create_gluon_model(num_features, num_classes):
# Create Model in Gluon
net = nn.HybridSequential()
net.add(nn.Dense(num_features, activation="relu"))
net.add(nn.Dense(1000, activation="relu"))
net.add(nn.Dense(num_classes))
#net.hybridize()
net.initialize(init=init.Xavier(), ctx=mx.cpu())
return net
入力データの形状: (32, 20)
出力形状: (32, 4)
ラベル形状: (32, 4) num_classes = 4
トレーニングしようとすると：
def train_vmhnet(net, train_data_loader, valid_data_loader, batch_size=32):
criterion = gluon.loss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(), "sgd", {"learning_rate": 0.1})
# Start the training.
for epoch in range(1):
train_loss, train_acc, valid_acc = 0.0, 0.0, 0.0
tic = time.time()
for batch_idx, (data, label) in enumerate(train_data_loader):
data = data.as_in_context(mx.cpu(0))
label = label.as_in_context(ctx)
# forward + backward
with autograd.record():
output = net(data)
loss = criterion(output, label)
loss.backward()
# update parameters
trainer.step(data.shape[0])
# calculate training metrics
train_loss += loss.mean().asscalar()
train_acc += acc(output, label)
print(epoch)
# calculate validation accuracy
for batch_idx, (data, label) in enumerate(valid_data_loader):
data = data.as_in_context(mx.cpu(0))
valid_acc += acc(net(data), label)
print(
"Epoch %d: loss %.3f, train acc %.3f, test acc %.3f, in %.1f sec"
% (
epoch,
train_loss / len(dataset_train),
train_acc / len(dataset_train),
valid_acc / len(dataset_test),
time.time() - tic,
)
)
IM で次のエラーが発生します。
mxnet.base.MXNetError: Shape inconsistent, Provided = [32,4], inferred shape=[32,1]
助けてください