search

npow / ubottu

Next Utterance Classification
http://arxiv.org/abs/1506.08909
136 stars 45 forks source link

it always stuck at 16% #6

Closed andyyuan78 closed 7 years ago

andyyuan78 commented 8 years ago

--2016-03-12 20:24:06-- (try:13) http://219.233.14.49/cs.mcgill.ca/~npow1/data/ubuntu_blobs.tgz Connecting to 219.233.14.49:80... connected. HTTP request sent, awaiting response... 206 Partial Content Length: 538404774 (513M), 451635110 (431M) remaining [application/x-gzip] Saving to: ‘ubuntu_blobs.tgz’

16% [++++++++++++++++++++++ ] 86,769,664 --.-K/s

npow commented 8 years ago

Try this link: http://cs.mcgill.ca/~jpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz

andyyuan78 commented 8 years ago

even worse;(

envy@ub1404:~/os_pri/github/ubottu$ wget -c http://cs.mcgill.ca/%7Ejpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz --2016-03-13 12:18:13-- http://cs.mcgill.ca/%7Ejpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz Resolving cs.mcgill.ca (cs.mcgill.ca)... 132.206.51.10 Connecting to cs.mcgill.ca (cs.mcgill.ca)|132.206.51.10|:80... connected. HTTP request sent, awaiting response... 302 Found Location: http://219.233.14.49/cs.mcgill.ca/%7Ejpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz [following] --2016-03-13 12:18:14-- http://219.233.14.49/cs.mcgill.ca/%7Ejpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz Connecting to 219.233.14.49:80... connected. HTTP request sent, awaiting response... 200 OK Length: unspecified [text/html] Saving to: ‘ubuntu_blobs.tgz’

[        <=>                                                                                                                                   ] 163,669     80.0KB/s   in 2.0s

2016-03-13 12:18:19 (80.0 KB/s) - ‘ubuntu_blobs.tgz’ saved [163669]

envy@ub1404:~/os_pri/github/ubottu$ ll total 184 drwxrwxr-x 4 envy envy 4096 3月 13 12:18 ./ drwxrwxr-x 27 envy envy 4096 3月 12 20:33 ../ drwxrwxr-x 8 envy envy 4096 3月 12 20:02 .git/ -rw-rw-r-- 1 envy envy 127 3月 12 20:02 .gitignore -rw-rw-r-- 1 envy envy 901 3月 12 20:02 README.md drwxrwxr-x 2 envy envy 4096 3月 12 20:02 src/ -rw-rw-r-- 1 envy envy 163669 3月 13 12:18 ubuntu_blobs.tgz envy@ub1404:~/os_pri/github/ubottu$

npow commented 8 years ago

Are you going through a proxy? On Mar 12, 2016 8:24 PM, "Andy Yuan" notifications@github.com wrote:

even worse;(

envy@ub1404:~/os_pri/github/ubottu$ wget -c http://cs.mcgill.ca/%7Ejpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz --2016-03-13 12:18:13-- http://cs.mcgill.ca/%7Ejpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz Resolving cs.mcgill.ca (cs.mcgill.ca)... 132.206.51.10 Connecting to cs.mcgill.ca (cs.mcgill.ca)|132.206.51.10|:80... connected. HTTP request sent, awaiting response... 302 Found Location: http://219.233.14.49/cs.mcgill.ca/%7Ejpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz [following] --2016-03-13 12:18:14-- http://219.233.14.49/cs.mcgill.ca/%7Ejpineau/datasets/ubuntu-corpus-1.0/ubuntu_blobs.tgz Connecting to 219.233.14.49:80... connected. HTTP request sent, awaiting response... 200 OK Length: unspecified [text/html] Saving to: ‘ubuntu_blobs.tgz’

[ <=> ] 163,669 80.0KB/s in 2.0s

2016-03-13 12:18:19 (80.0 KB/s) - ‘ubuntu_blobs.tgz’ saved [163669]

envy@ub1404:~/os_pri/github/ubottu$ ll total 184 drwxrwxr-x 4 envy envy 4096 3月 13 12:18 ./ drwxrwxr-x 27 envy envy 4096 3月 12 20:33 ../ drwxrwxr-x 8 envy envy 4096 3月 12 20:02 .git/ -rw-rw-r-- 1 envy envy 127 3月 12 20:02 .gitignore -rw-rw-r-- 1 envy envy 901 3月 12 20:02 README.md drwxrwxr-x 2 envy envy 4096 3月 12 20:02 src/ -rw-rw-r-- 1 envy envy 163669 3月 13 12:18 ubuntu_blobs.tgz envy@ub1404:~/os_pri/github/ubottu$

— Reply to this email directly or view it on GitHub https://github.com/npow/ubottu/issues/6#issuecomment-195868881.

andyyuan78 commented 8 years ago

no proxy

basma-b commented 7 years ago

Hey @npow, I am having a problem on running the code with LSTM and RNN :

TypeError: ('Bad input argument to theano function with name "main.py:133" at index 0(0-based)', 'TensorType(float32, vector) cannot store a value of dtype float64 without risking loss of precision. If you do not mind this loss, you can: 1) explicitly cast your data to float32, or 2) set "allow_input_downcast=True" when calling "function".', array([ 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., ......

After a Google search I understood that the problem is caused by a theano's type, it could be solved by adding a cast to np.int32 but I can't see where I can place it in the code.

npow commented 7 years ago

Did you try setting floatX=float32 in the theano config?

On Nov 25, 2016 12:22 PM, "Basma BOUSSAHA" notifications@github.com wrote:

Hey @npow https://github.com/npow, I am having a problem on running the code with LSTM and RNN :

TypeError: ('Bad input argument to theano function with name "main.py:133" at index 0(0-based)', 'TensorType(float32, vector) cannot store a value of dtype float64 without risking loss of precision. If you do not mind this loss, you can: 1) explicitly cast your data to float32, or 2) set "allow_input_downcast=True" when calling "function".', array([ 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., ......

After a Google search I understood that the problem is caused by a theano's type, it could be solved by adding a cast to np.int32 but I can't see where I can place it in the code.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/npow/ubottu/issues/6#issuecomment-262978199, or mute the thread https://github.com/notifications/unsubscribe-auth/AAMleWhBz4UNEb_hCkwl0HP6Tqn1Da_wks5rBv0tgaJpZM4HvSq4 .

basma-b commented 7 years ago

Hi Nissan,

I am Basma a PhD student at the university of Nantes (France) and I am working on Next Utterance Classification. Actually I am working on the Ubuntu Dialogue Corpus and I am trying to use some different classification methods on this corpus. I would like to compare my results to those you report on Github using RNN and LSTM and I want to have the scores you get for each answer in the test set. Due to some computational limitation (absence of GPU on our servers) unfortunately I am not able to run your code and get individual scores for each response.

I wish if you could send them to me for both RNN and LSTM and I would be very thankful.

Many thanks, Basma

On 26 November 2016 at 11:09, Nissan Pow notifications@github.com wrote:

Did you try setting floatX=float32 in the theano config?

On Nov 25, 2016 12:22 PM, "Basma BOUSSAHA" notifications@github.com wrote:

Hey @npow https://github.com/npow, I am having a problem on running the code with LSTM and RNN :

TypeError: ('Bad input argument to theano function with name "main.py:133" at index 0(0-based)', 'TensorType(float32, vector) cannot store a value of dtype float64 without risking loss of precision. If you do not mind this loss, you can: 1) explicitly cast your data to float32, or 2) set "allow_input_downcast=True" when calling "function".', array([ 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., ......

After a Google search I understood that the problem is caused by a theano's type, it could be solved by adding a cast to np.int32 but I can't see where I can place it in the code.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/npow/ubottu/issues/6#issuecomment-262978199, or mute the thread https://github.com/notifications/unsubscribe-auth/AAMleWhBz4UNEb_ hCkwl0HP6Tqn1Da_wks5rBv0tgaJpZM4HvSq4 .

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/npow/ubottu/issues/6#issuecomment-263055221, or mute the thread https://github.com/notifications/unsubscribe-auth/AJfNkEKBy3EbFb674UEiQiq35cZ0DW_Wks5rCAVxgaJpZM4HvSq4 .

npow commented 7 years ago

@basma-b I don't have the dump of all the test probabilities anymore, but the following links has the probabilities for all the test examples which were incorrectly classified, and 35k of the correct classifications.

http://cs.mcgill.ca/~npow1/data/ubuntu_errors.html http://cs.mcgill.ca/~npow1/data/ubuntu_correct.html