Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Source/Target length distinction #1

Merged
merged 2 commits into from
Feb 23, 2017
Merged

Source/Target length distinction #1

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
acbdbd9
Splitting preprocess.py parameters -seq_length
pltrdy Feb 23, 2017
e5b5526
Training: printing src token/sec instead of tgt token/sec
pltrdy Feb 23, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 7 additions & 5 deletions preprocess.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,8 +34,10 @@
help="Path to an existing target vocabulary")


parser.add_argument('-seq_length', type=int, default=50,
help="Maximum sequence length")
parser.add_argument('-src_seq_length', type=int, default=50,
help="Maximum source sequence length")
parser.add_argument('-tgt_seq_length', type=int, default=50,
help="Maximum target sequence length")
parser.add_argument('-shuffle', type=int, default=1,
help="Shuffle data")
parser.add_argument('-seed', type=int, default=3435,
Expand Down Expand Up @@ -108,7 +110,7 @@ def makeData(srcFile, tgtFile, srcDicts, tgtDicts):
print('WARNING: source and target do not have the same number of sentences')
break

if len(srcWords) <= opt.seq_length and len(tgtWords) <= opt.seq_length:
if len(srcWords) <= opt.src_seq_length and len(tgtWords) <= opt.tgt_seq_length:

src += [srcDicts.convertToIdx(srcWords,
onmt.Constants.UNK_WORD)]
Expand Down Expand Up @@ -141,8 +143,8 @@ def makeData(srcFile, tgtFile, srcDicts, tgtDicts):
src = [src[idx] for idx in perm]
tgt = [tgt[idx] for idx in perm]

print('Prepared %d sentences (%d ignored due to length == 0 or > %d)' %
(len(src), ignored, opt.seq_length))
print('Prepared %d sentences (%d ignored due to length == 0 or source length > %d or target length > %d)' %
(len(src), ignored, opt.src_seq_length, opt.tgt_seq_length))

return src, tgt

Expand Down
8 changes: 5 additions & 3 deletions train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -164,6 +164,7 @@ def trainEpoch(epoch):

total_loss, report_loss = 0, 0
total_words, report_words = 0, 0
report_src_words = 0
start = time.time()
for i in range(len(trainData)):

Expand All @@ -183,16 +184,17 @@ def trainEpoch(epoch):

report_loss += loss
total_loss += loss
report_src_words += batch[0].data.ne(onmt.Constants.PAD).sum()
num_words = targets.data.ne(onmt.Constants.PAD).sum()
total_words += num_words
report_words += num_words
if i % opt.log_interval == 0 and i > 0:
print("Epoch %2d, %5d/%5d batches; perplexity: %6.2f; %3.0f tokens/s" %
print("Epoch %2d, %5d/%5d batches; perplexity: %6.2f; %3.0f Source tokens/s" %
(epoch, i, len(trainData),
math.exp(report_loss / report_words),
report_words/(time.time()-start)))
report_src_words/(time.time()-start)))

report_loss = report_words = 0
report_loss = report_words = report_src_words = 0
start = time.time()

return total_loss / total_words
Expand Down