The test loss does not change

[aalesar]

After fixing this error:
Traceback (most recent call last):
File "train.py", line 344, in
exp.train()
File "train.py", line 327, in train
loss = self.train_batch(i)
File "train.py", line 304, in train_batch
coeff = clip_gradient(self.mdl, self.args.clip_norm)
File "train.py", line 36, in clip_gradient
modulenorm = p.grad.data.norm()
AttributeError: 'NoneType' object has no attribute 'data',
by checking if p.grad is not None before ,
the test loss would change only slightly, but still is 0.901449918747 after 30 epochs,
also I'm using cpu and not gpu. Do you have any thoughts?