search

shlurbee / dmrs-text-generation-naacl2019

Code in support of our NAACL 2019 short paper "Neural Text Generation from Rich Semantic Representations"
1 stars 4 forks source link

Reproducing experiment results of the paper #3

Open aaronlolo326 opened 5 years ago

aaronlolo326 commented 5 years ago

Hello @shlurbee! I tried reproducing the results of the paper. However, after following the first few steps, the following error occurred.

bash-4.2$ sh scripts/run\_experiments.sh
Usage: From root dir run:
sh scripts/run_experiments.sh
rm: cannot remove ‘models/v13*’: No such file or directory
rm: cannot remove ‘models/v12*’: No such file or directory
Traceback (most recent call last):
  File "OpenNMT-py/train.py", line 503, in <module>
    main()
...
FileNotFoundError: [Errno 2] No such file or directory: 'data/opennmt.train.pt'

As there are quite a number of run_*.sh files under script/, may you give us some instructions or the exact steps to produce all the results of your work? Thanks in advance!

shlurbee commented 5 years ago

Hi! Thanks for your interest in our research. To train only the best-performing model from the paper, go to the bottom of scripts/run_silver3.sh and insert # to comment out the commands on lines 179-236. So, comment out everything highlighted here: https://github.com/shlurbee/dmrs-text-generation-naacl2019/blob/master/scripts/run_silver3.sh#L179-L236

Then, uncomment just the following lines: prep_data preprocess_vs 6 850000 run_data_checks

Run the file (sh scripts/run_silver3.sh) and make sure the numbers in the output look correct. (They will say something like "expected about X number of lines, found X number of lines", and the numbers should be the same order of magnitude). Unfortunately we didn't set up the scripts to stop right away when there's a problem, hence running a few lines at a time.

If everything looked ok, comment out the above lines again and uncomment this one: train_vs data_vs6 2 You may want to change that final 2 to a 0 if your machine has fewer than 3 gpus for training. Then run the script again to start training. It took our machine about 43 hours to train this model for 22 epochs.

On Tue, Oct 8, 2019 at 6:27 AM aaronlolo326 notifications@github.com wrote:

Hello @shlurbee https://github.com/shlurbee! I tried reproducing the results of the paper. However, after following the first few steps, the following error occurred.

bash-4.2$ sh scripts/run_experiments.sh

Usage: From root dir run:

sh scripts/run_experiments.sh

rm: cannot remove ‘models/v13*’: No such file or directory

rm: cannot remove ‘models/v12*’: No such file or directory

Traceback (most recent call last):

File "OpenNMT-py/train.py", line 503, in 

main()

...

FileNotFoundError: [Errno 2] No such file or directory: 'data/opennmt.train.pt'

As there are quite a number of run_*.sh files under script/, may you give us some instructions or the exact steps to produce all the results of your work? Thanks in advance!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/shlurbee/dmrs-text-generation-naacl2019/issues/3?email_source=notifications&email_token=AACAYK7EYRPLY4S5MX2NNHTQNSDE3A5CNFSM4I6RXT4KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HQLDIBQ, or mute the thread https://github.com/notifications/unsubscribe-auth/AACAYK3MVUL75IRPA7M6FCLQNSDE3ANCNFSM4I6RXT4A .