def test_opening_two_instances(self, writer_stock):
with pytest.raises(PermissionError):
StockRoom(enable_write=True)
arr = np.arange(20).reshape(4, 5)
oldarr = arr * randint(1, 100)
col1 = writer_stock.data["ndcol"]
col1[1] = oldarr
writer_stock.commit("added data")
stock2 = StockRoom()
col2 = stock2.data["ndcol"]
assert np.allclose(col2[1], oldarr)
with pytest.raises(PermissionError):
with stock2.enable_write():
pass
stock2._repo._env._close_environments()
optimizer = optim.SGD(net.parameters(), lr=lr, momentum=momentum)
for epoch in range(2):
running_loss = 0.0
current_loss = 99999
best_loss = 99999
p = tqdm(dloader)
p.set_description('[epcoh: %d, iteration: %d] loss: %5d' %
(epoch + 1, 1, current_loss))
for i, data in enumerate(p):
inputs, labels = data
optimizer.zero_grad()
outputs = net(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
running_loss += loss.item()
if i % check_every == check_every - 1:
current_loss = running_loss / check_every
running_loss = 0.0
p.set_description('[epcoh: %d, iteration: %d] loss: %.6f' %
(epoch + 1, i + 1, current_loss))
if current_loss < best_loss:
with stock.enable_write():
stock.experiment['lr'] = lr
stock.experiment['momentum'] = momentum
stock.model['cifarmodel'] = net.state_dict()
best_loss = current_loss
print(stock.model.keys())
print('Finished Training')