rnnlm adaptation by the cache model on swbd #2659
Conversation
There was a problem hiding this comment.
@keli78
Please merge my pull request in your repository firstly. I corrected some simple mistakes conveniently, such as line is longer than 80, missing spaces, and so on.
Please keep in mind, you always can use misc/maintenance/cpplint.py to do some simple coding style check.
Otherwise, I added a few comments.
| }; | ||
| } // namespace kaldi | ||
|
|
||
| void ReadUttToConvo(string filename, map<string, string> &m) { |
There was a problem hiding this comment.
please use pointer instead of reference. When you use a reference as the input argument, please use const reference.
Otherwise, in kaldi, when you use string, please use std::string.
| if ((i + idx) >= utt_ids.size()) { | ||
| idx -= range; | ||
| } | ||
| KALDI_ASSERT(idx < int(utt_ids.size())); |
There was a problem hiding this comment.
Please use C++ casting style. Use static_cast() instead.
| }; | ||
| } // namespace kaldi | ||
|
|
||
| void ReadUttToConvo(string filename, map<string, string> &m) { |
There was a problem hiding this comment.
Please use pointer instead of non-const reference.
Otherwise, in kaldi, when you use string, please use std::string.
| #include "lat/kaldi-lattice.h" | ||
| #include "lat/lattice-functions.h" | ||
| #include "lat/compose-lattice-pruned.h" | ||
| #include <fstream> |
There was a problem hiding this comment.
In most cases, the order of header should be: current file header, c system header, c++ system header, other
| #include "lat/kaldi-lattice.h" | ||
| #include "lat/lattice-functions.h" | ||
| #include "lat/compose-lattice-pruned.h" | ||
| #include <fstream> |
| } // namespace kaldi | ||
|
|
||
| // read the file and genereate a map from [utt-id] to [convo-id], stored in *m | ||
| void ReadUttToConvo(string filename, map<string, string> &m) { |
| using std::unordered_map; | ||
|
|
||
| // read the file and genereate a map from [utt-id] to [convo-id], stored in *m | ||
| void ReadUttToConvo(string filename, map<string, string> &m) { |
There was a problem hiding this comment.
The same as above.
I think this function is used everywhere. Could you put it into a public head file?
| @@ -0,0 +1,245 @@ | |||
| // rnnlmbin/rnnlm-compute-prob.cc | |||
There was a problem hiding this comment.
wrong. Do you think this is a meaningful name "adjust2"?
| // std::cout << rnnlm_compute_state.LogProbOfWord(word_id) << " "; | ||
| CuMatrix<BaseFloat> word_logprobs(1, word_embedding_mat.NumRows()); | ||
| rnnlm_compute_state.GetLogProbOfWords(&word_logprobs); | ||
| // std::cout << word_logprobs(0, word_id) << " "; // this should be exactly the same as the 3 lines above |
There was a problem hiding this comment.
If this part of code will not use, you'd better delete them
| using std::unordered_map; | ||
|
|
||
| // read the file and genereate a map from [utt-id] to [convo-id], stored in *m | ||
| void ReadUttToConvo(string filename, map<string, string> &m) { |
| "Usage: lattice-lmrescore-kaldi-rnnlm-pruned-cache-adapt-onlyhistory [options] \\\n" | ||
| " <old-lm-rxfilename> <embedding-file> \\\n" | ||
| " <raw-rnnlm-rxfilename> \\\n" | ||
| " <lattice-rspecifier> <lattice-wspecifier>\n" |
There was a problem hiding this comment.
I notice the usage message of this program does not match the actual usage. Could you please clean it up a bit?
There was a problem hiding this comment.
@danpovey I have modified the usage part and refactored the code.
Ke
…/o history (to estimate unigram counts)
|
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. |
|
This issue has been automatically closed by a bot strictly because of inactivity. This does not mean that we think that this issue is not important! If you believe it has been closed hastily, add a comment to the issue and mention @kkm000, and I'll gladly reopen it. |
|
This issue has been automatically marked as stale by a bot solely because it has not had recent activity. Please add any comment (simply 'ping' is enough) to prevent the issue from being closed for 60 more days if you believe it should be kept open. |
|
Hello, I've been checking this dissertation: https://jscholarship.library.jhu.edu/handle/1774.2/66812 Any info would be great, thanks. |
|
It looks to me like it might have been an oversight on my part. |
|
Thanks, I'll check. |
|
This issue has been automatically marked as stale by a bot solely because it has not had recent activity. Please add any comment (simply 'ping' is enough) to prevent the issue from being closed for 60 more days if you believe it should be kept open. |
Rnnlm adaptation by the cache model on swbd (no weighting trick)