Difference between revisions of "2013-04-26"

From cslt Wiki
Jump to: navigation, search
Line 29: Line 29:
 
*Our results are with 400 hour AM, 88k LM. ML+bMMI.
 
*Our results are with 400 hour AM, 88k LM. ML+bMMI.
 
*The CSLT structure: 300*[1200*1200*1200*40*1200]*4850.  
 
*The CSLT structure: 300*[1200*1200*1200*40*1200]*4850.  
 
+
*The CSLT feature: MFCC+delta MFCC
 
* compare with the traditional structure 300*[1200*1200*1200*1200*1200]*4850.
 
* compare with the traditional structure 300*[1200*1200*1200*1200*1200]*4850.
 
  
 
===Tencent test result===
 
===Tencent test result===
  
:  AM: 70h training data(2 day, 15 machines, 10 threads)
+
:  AM: 70h training data
 
:  LM: 88k LM  
 
:  LM: 88k LM  
 
:  Test case: general  
 
:  Test case: general  
  
 
{|class="wikitable"
 
{|class="wikitable"
!Feature !! GMM-bMMI !! DNN !! DNN-MMI
+
!Feature !! GMM !!GMM-bMMI !! DNN !! DNN-MMI
 
|-
 
|-
|PLP(-5,+5) || 38.4  || 26.5 || 23.8  
+
|PLP(-5,+5) [Eryu]          || 47 || 38.4  || 26.5 || 23.8  
 
|-
 
|-
|PLP+LDA+MLLT(-5,+5) || 38.4 ||28.7
+
|PLP+LDA+MLLT(-5,+5)[Jingbo] || 47 || -      || 34
 
|-
 
|-
 
|}
 
|}
 +
 +
* Tencent NN structure:
 +
:300*[1200*1200*1200*1200]*1700, #param=700k
 +
:300*[1007*1007*1007*1007]*3xxx  #param=700k
 +
 +
:*CSLT reproduce phone-clustered based NN
 +
:*CSLT investigate performance of different epochs.
  
  
 
===GPU & CPU merge===
 
===GPU & CPU merge===
: Invesigate the possibility to merge GPU and CPU code. GPU computing code merged to CPU.  
+
: Investigate the possibility to merge GPU and CPU code.  
 +
: CUDA code merged to CPU.  
  
 
===L-1 sparse initial training===
 
===L-1 sparse initial training===

Revision as of 06:20, 26 April 2013

Data sharing

  • LM count files are still in transfering.

DNN progress

400 hour DNN training

Test Set Tencent Baseline bMMI fMMI BN(with fMMI) Hybrid
1900 8.4 7.65 7.35 6.57 7.27
2044 22.4 24.44 24.03 21.77 20.24
online1 35.6 34.66 34.33 31.44 30.53
online2 29.6 27.23 26.80 24.10 23.89
map 24.5 27.54 27.69 23.79 22.46
notepad 16 19.81 21.75 15.81 12.74
general 36 38.52 38.90 33.61 31.55
speedup 26.8 27.88 26.81 22.82 22.00
  • Tencent baseline is with 700h online data+ 700h 863 data, HLDA+MPE, 88k lexicon
  • Our results are with 400 hour AM, 88k LM. ML+bMMI.
  • The CSLT structure: 300*[1200*1200*1200*40*1200]*4850.
  • The CSLT feature: MFCC+delta MFCC
  • compare with the traditional structure 300*[1200*1200*1200*1200*1200]*4850.

Tencent test result

AM: 70h training data
LM: 88k LM
Test case: general
Feature GMM GMM-bMMI DNN DNN-MMI
PLP(-5,+5) [Eryu] 47 38.4 26.5 23.8
PLP+LDA+MLLT(-5,+5)[Jingbo] 47 - 34
  • Tencent NN structure:
300*[1200*1200*1200*1200]*1700, #param=700k
300*[1007*1007*1007*1007]*3xxx #param=700k
  • CSLT reproduce phone-clustered based NN
  • CSLT investigate performance of different epochs.


GPU & CPU merge

Investigate the possibility to merge GPU and CPU code.
CUDA code merged to CPU.

L-1 sparse initial training

Start to investigating.

Kaldi/HTK merge

  • HTK2Kaldi: hold.
  • Kaldi2HTK: done with implementation. Performance improved.

Embedded progress

  • PocketSphinx migration done. Very slow.
  • QA LM training, done.