Fix python 3 download script and left-over cuda in attention layer
Browse files- scripts/download_weights.py +2 -1
- torchmoji/attlayer.py +0 -2
- torchmoji/finetuning.py +0 -3
scripts/download_weights.py
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
from __future__ import print_function
|
2 |
import os
|
3 |
from subprocess import call
|
|
|
4 |
|
5 |
curr_folder = os.path.basename(os.path.normpath(os.getcwd()))
|
6 |
|
@@ -23,7 +24,7 @@ def prompt():
|
|
23 |
'n': False,
|
24 |
'no': False,
|
25 |
}
|
26 |
-
choice =
|
27 |
if choice in valid:
|
28 |
return valid[choice]
|
29 |
else:
|
|
|
1 |
from __future__ import print_function
|
2 |
import os
|
3 |
from subprocess import call
|
4 |
+
from builtins import input
|
5 |
|
6 |
curr_folder = os.path.basename(os.path.normpath(os.getcwd()))
|
7 |
|
|
|
24 |
'n': False,
|
25 |
'no': False,
|
26 |
}
|
27 |
+
choice = input().lower()
|
28 |
if choice in valid:
|
29 |
return valid[choice]
|
30 |
else:
|
torchmoji/attlayer.py
CHANGED
@@ -51,8 +51,6 @@ class Attention(Module):
|
|
51 |
# See e.g. https://discuss.pytorch.org/t/self-attention-on-words-and-masking/5671/5
|
52 |
max_len = unnorm_ai.size(1)
|
53 |
idxes = torch.arange(0, max_len, out=torch.LongTensor(max_len)).unsqueeze(0)
|
54 |
-
if torch.cuda.is_available():
|
55 |
-
idxes = idxes.cuda()
|
56 |
mask = Variable((idxes < input_lengths.unsqueeze(1)).float())
|
57 |
|
58 |
# apply mask and renormalize attention scores (weights)
|
|
|
51 |
# See e.g. https://discuss.pytorch.org/t/self-attention-on-words-and-masking/5671/5
|
52 |
max_len = unnorm_ai.size(1)
|
53 |
idxes = torch.arange(0, max_len, out=torch.LongTensor(max_len)).unsqueeze(0)
|
|
|
|
|
54 |
mask = Variable((idxes < input_lengths.unsqueeze(1)).float())
|
55 |
|
56 |
# apply mask and renormalize attention scores (weights)
|
torchmoji/finetuning.py
CHANGED
@@ -514,9 +514,6 @@ def fit_model(model, loss_op, optim_op, train_gen, val_gen, epochs,
|
|
514 |
X_train, y_train = data
|
515 |
X_train = Variable(X_train, requires_grad=False)
|
516 |
y_train = Variable(y_train, requires_grad=False)
|
517 |
-
if torch.cuda.is_available():
|
518 |
-
X_train = X_train.cuda()
|
519 |
-
y_train = y_train.cuda()
|
520 |
model.train()
|
521 |
optim_op.zero_grad()
|
522 |
output = model(X_train)
|
|
|
514 |
X_train, y_train = data
|
515 |
X_train = Variable(X_train, requires_grad=False)
|
516 |
y_train = Variable(y_train, requires_grad=False)
|
|
|
|
|
|
|
517 |
model.train()
|
518 |
optim_op.zero_grad()
|
519 |
output = model(X_train)
|