dashayushman / deep-trans Goto Github PK

i have Nvidia 940mx 4gb when I run training on my own data (different languages)

if i use

--size=1024 --num_layers=5 --steps_per_checkpoint=1000

it stop in this line for more 15 minutes

I tensorflow/core/common_runtime/gpu/pool_allocator.cc:259] Raising pool_size_limit_ from 2059 to

when i use smaller size and less layers
--size=512 --num_layers=2 --steps_per_checkpoint=1000

the result is

global step 1000 learning rate 0.0010 step-time 0.19 perplexity 1861.61
  eval: empty bucket 0
  eval: bucket 1 perplexity 6.93
  eval: bucket 2 perplexity 6.45
  eval: empty bucket 3
  eval: empty bucket 4
  eval: empty bucket 5
global step 2000 learning rate 0.0010 step-time 0.20 perplexity 4.04
  eval: empty bucket 0
  eval: bucket 1 perplexity 2.84
  eval: bucket 2 perplexity 3.03
  eval: empty bucket 3
  eval: empty bucket 4
  eval: empty bucket 5
global step 3000 learning rate 0.0010 step-time 0.20 perplexity 2.27
  eval: empty bucket 0
  eval: bucket 1 perplexity 2.12
  eval: bucket 2 perplexity 2.09
  eval: empty bucket 3
  eval: empty bucket 4
  eval: empty bucket 5

and continue to more and more step take more than 2 hours and not finish

how ever it small data just few words.

Is it the number of words we provide or the number of characters in vocabulary file which is generated by the code.

Traceback (most recent call last):
File "transliterate.py", line 340, in main
decode()
File "transliterate.py", line 238, in decode
for output in outputs]))
IndexError: list index out of range

Traceback (most recent call last):
File "transliterate.py", line 32, in
import seq2seq_model
File "/data/home/kunal/Desktop/ML-Project/deep-trans/seq2seq_model.py", line 14, in
from tensorflow.models.rnn.translate import data_utils
ImportError: No module named models.rnn.translate

These 2 datasets (train.rel.2.en & train.rel.2.hn), which are needed for training the model, are not in the source directory to download!
How can I download Them?
Should I prepare them by myself

While compiling the transliterate.py file,it is showing many errors.
Without removing errors, I cannot go to the next step. As you said that this file should be compiled without any errors.
Is it necessary to compile this file successfully?
So please reply as soon as possible.