Tdnn pool

.gitattributes

@@ -15,4 +15,6 @@ windows/INSTALL*   eol=native
 windows/NewGuidCmd.exe.config text eol=crlf
 windows/NewGuidCmd.exe binary
 
+# Prevent git changing CR-LF to LF when archiving (patch requires CR-LF on Windows).
+**/*.patch            -text
 

.gitignore

@@ -6,11 +6,12 @@
 !/src/*/Makefile
 !/src/*/README
 
-# Compiled Object files
+# Compiled Object files and python ciles
 *.slo
 *.lo
 *.o
 *.obj
+*.pyc
 
 # Compiled Dynamic libraries
 *.so
@@ -81,6 +82,8 @@ GSYMS
 /tools/portaudio/
 /tools/sctk-2.4.0-20091110-0958.tar.bz2
 /tools/sctk-2.4.0/
+/tools/sctk-2.4.10-20151007-1312Z.tar.bz2
+/tools/sctk-2.4.10/
 /tools/sph2pipe_v2.5.tar.gz
 /tools/sph2pipe_v2.5/
 /tools/kaldi_lm.tar.gz
@@ -98,5 +101,9 @@ GSYMS
 /tools/mpg123
 /tools/mpg123-1.21.0.tar.bz2
 /tools/mpg123-1.21.0
+/tools/pthreads
+/tools/pthreads*.zip
 /tools/sequitur
 /tools/srilm.tgz
+
+/kaldiwin_vs*

egs/ami/s5/RESULTS_ihm

@@ -6,6 +6,9 @@ for x in exp/ihm/{mono,tri,sgmm,nnet,dnn,lstm}*/decode*; do [ -d $x ] && [[ $x =
 
 exit 0
 
+# Results with close-talk microphones (IHM),
+
+# Pawel,
 dev
 exp/ihm/tri2a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_13/dev.ctm.filt.dtl:Percent Total Error       =   38.0%   (35925)
 exp/ihm/tri3a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_14/dev.ctm.filt.dtl:Percent Total Error       =   35.3%   (33329)
@@ -18,14 +21,58 @@ exp/ihm/tri3a/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_14/eval.ctm.filt.dtl:Perce
 exp/ihm/tri4a/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_13/eval_o4.ctm.filt.dtl:Percent Total Error       =   35.0%   (31463)
 exp/ihm/tri4a_mmi_b0.1/decode_eval_3.mdl_ami_fsh.o3g.kn.pr1-7/ascore_12/eval_o4.ctm.filt.dtl:Percent Total Error       =   31.7%   (28518)
 
+# Karel, JSALT 2015, (21.7.2015)
+
+# dev,
+## GMM,
+%WER 38.1 | 13098 94489 | 67.1 20.6 12.2 5.2 38.1 67.0 | exp/ihm/tri2a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_13/dev.ctm.filt.sys
+%WER 35.5 | 13098 94487 | 69.6 19.0 11.4 5.1 35.5 65.8 | exp/ihm/tri3a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_13/dev.ctm.filt.sys
+%WER 32.2 | 13098 94483 | 72.5 17.2 10.3 4.8 32.2 63.8 | exp/ihm/tri4a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_14/dev.ctm.filt.sys #0.1% worse than Pawel!
+%WER 30.2 | 13098 94479 | 74.0 15.6 10.4 4.2 30.2 61.9 | exp/ihm/tri4a_mmi_b0.1/decode_dev_3.mdl_ami_fsh.o3g.kn.pr1-7/ascore_12/dev.ctm.filt.sys
+## DNN-Xent,
+%WER 26.0 | 13098 94483 | 77.9 13.5 8.5 4.0 26.0 58.4 | exp/ihm/dnn4_pretrain-dbn_dnn/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_11/dev.ctm.filt.sys
+## DNN-sMBR,
+%WER 24.9 | 13098 94484 | 79.2 13.2 7.6 4.1 24.9 57.1 | exp/ihm/dnn4_pretrain-dbn_dnn_smbr/decode_dev_ami_fsh.o3g.kn.pr1-7_it1/ascore_11/dev.ctm.filt.sys
+%WER 24.3 | 13098 94481 | 79.6 12.6 7.8 3.9 24.3 56.3 | exp/ihm/dnn4_pretrain-dbn_dnn_smbr/decode_dev_ami_fsh.o3g.kn.pr1-7_it4/ascore_12/dev.ctm.filt.sys
+
+# eval,
+## GMM,
+%WER 43.9 | 12643 89978 | 60.8 25.3 13.9 4.8 43.9 65.6 | exp/ihm/tri2a/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_14/eval.ctm.filt.sys
+%WER 40.8 | 12643 89985 | 63.8 23.6 12.6 4.7 40.8 64.6 | exp/ihm/tri3a/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_14/eval.ctm.filt.sys
+%WER 35.1 | 12643 89975 | 69.1 19.8 11.1 4.2 35.1 61.8 | exp/ihm/tri4a/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_14/eval.ctm.filt.sys #0.1% worse than Pawel!
+%WER 31.7 | 12643 89986 | 72.1 18.0 9.9 3.8 31.7 59.4 | exp/ihm/tri4a_mmi_b0.1/decode_eval_4.mdl_ami_fsh.o3g.kn.pr1-7/ascore_11/eval.ctm.filt.sys
+## DNN-Xent,
+%WER 27.1 | 12643 89971 | 76.4 15.5 8.1 3.5 27.1 57.2 | exp/ihm/dnn4_pretrain-dbn_dnn/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_10/eval.ctm.filt.sys
+## DNN-sMBR,
+%WER 25.4 | 12643 89974 | 77.9 14.7 7.4 3.3 25.4 55.1 | exp/ihm/dnn4_pretrain-dbn_dnn_smbr/decode_eval_ami_fsh.o3g.kn.pr1-7_it1/ascore_11/eval.ctm.filt.sys
+%WER 24.6 | 12643 89972 | 78.8 14.1 7.1 3.3 24.6 54.4 | exp/ihm/dnn4_pretrain-dbn_dnn_smbr/decode_eval_ami_fsh.o3g.kn.pr1-7_it4/ascore_11/eval.ctm.filt.sys
+
 
-# TDNN results
+# Vijay, TDNN results,
 for x in exp/$mic/nnet2_online/*/decode*; do [ -d $x ] && [[ $x =~ "$1" ]] && grep Sum $x/ascore_*/*.sys | utils/best_wer.sh; done
-#dev
+# dev,
 %WER 25.0 | 13098 94483 | 78.3 12.0 9.6 3.4 25.0 57.7 | exp/ihm/nnet2_online/nnet_ms_sp_online/decode_dev/ascore_13/dev_hires.ctm.filt.sys
 %WER 25.3 | 13098 94468 | 78.5 12.7 8.8 3.8 25.3 57.9 | exp/ihm/nnet2_online/nnet_ms_sp_online/decode_dev_utt/ascore_12/dev_hires.ctm.filt.sys
 %WER 25.0 | 13098 94476 | 78.5 12.4 9.1 3.6 25.0 58.0 | exp/ihm/nnet2_online/nnet_ms_sp_online/decode_dev_utt_offline/ascore_13/dev_hires.ctm.filt.sys
-#eval
+# eval,
 %WER 25.9 | 12643 89971 | 77.2 14.2 8.6 3.2 25.9 56.4 | exp/ihm/nnet2_online/nnet_ms_sp_online/decode_eval/ascore_12/eval_hires.ctm.filt.sys
 %WER 26.0 | 12643 89976 | 77.1 14.7 8.2 3.2 26.0 55.7 | exp/ihm/nnet2_online/nnet_ms_sp_online/decode_eval_utt/ascore_12/eval_hires.ctm.filt.sys
 %WER 25.8 | 12643 89978 | 77.6 14.6 7.8 3.4 25.8 55.8 | exp/ihm/nnet2_online/nnet_ms_sp_online/decode_eval_utt_offline/ascore_11/eval_hires.ctm.filt.sys
+
+#------------------------------------------------------------------------------------------------------------------------------------
+# Nnet3 systems
+
+# BLSTM
+#  local/nnet3/run_blstm.sh --mic ihm \
+#    --chunk-right-context 20 \
+#    --use-sat-alignments true
+# Note: Chunk right context of 20 limits the latency of the acoustic model to
+# 20 frames.
+
+%WER 22.8 | 13098 94494 | 80.1 11.0 8.9 3.0 22.8 54.8 | exp/ihm/nnet3/lstm_bidirectional_ld0/decode_dev/ascore_10/dev_hires.ctm.filt.sys
+%WER 22.6 | 12643 89969 | 80.0 12.7 7.3 2.7 22.6 53.5 | exp/ihm/nnet3/lstm_bidirectional_ld0/decode_eval/ascore_9/eval_hires.ctm.filt.sys
+
+## Chain systems
+  # local/chain/run_tdnn_ami_5.sh  --mic ihm --max-wer 50 --affix min_seg_len2_50wer (built with min-seg-len 2 secs, but script now just supports (frames_per_eg+5)/100)
+  %WER 22.4 | 13098 94484 | 80.5 10.7 8.8 3.0 22.4 54.8 | 0.091 | exp/ihm/chain/tdnn_min_seg_len2_50wer_sp/decode_dev/ascore_10/dev_hires.ctm.filt.sys
+  %WER 22.4 | 12643 89973 | 80.3 12.6 7.1 2.8 22.4 53.2 | 0.155 | exp/ihm/chain/tdnn_min_seg_len2_50wer_sp/decode_eval/ascore_10/eval_hires.ctm.filt.sys

egs/ami/s5/RESULTS_mdm

@@ -6,8 +6,9 @@ for x in exp/mdm*/{mono,tri,sgmm,nnet,dnn,lstm}*/decode*; do [ -d $x ] && [[ $x
 exit 0
 
 
-#Beamforming of 8 microphones, WER scores with up to 4 overlapping speakers
+# Beamforming of 8 microphones, WER scores with up to 4 overlapping speakers,
 
+# Pawel,
 dev
 exp/mdm8/tri2a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_13/dev_o4.ctm.filt.dtl:Percent Total Error       =   58.8%   (55568)
 exp/mdm8/tri3a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_13/dev_o4.ctm.filt.dtl:Percent Total Error       =   57.0%   (53855)
@@ -35,7 +36,7 @@ for x in exp/$mic/nnet2_online/*/decode*; do [ -d $x ] && [[ $x =~ "$1" ]] && gr
   #dev
     %WER 40.9 | 15965 94490 | 64.6 19.9 15.5 5.5 40.9 61.9 | -26.104 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch4_dev_utt/ascore_12/dev_hires_o4.ctm.filt.sys
     %WER 40.7 | 13961 94495 | 64.4 18.8 16.8 5.0 40.7 70.4 | -26.622 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch4_dev_utt_offline/ascore_13/dev_hires_o4.ctm.filt.sys
-  #eval  
+  #eval
     %WER 44.2 | 13577 89767 | 61.1 22.3 16.6 5.3 44.2 68.9 | -25.003 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch4_eval_utt/ascore_11/eval_hires_o4.ctm.filt.sys
     %WER 44.0 | 13448 89769 | 60.8 21.4 17.8 4.9 44.0 69.6 | -25.331 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch4_eval_utt_offline/ascore_12/eval_hires_o4.ctm.filt.sys
 
@@ -57,8 +58,8 @@ for x in exp/$mic/nnet2_online/*/decode*; do [ -d $x ] && [[ $x =~ "$1" ]] && gr
     %WER 40.9 | 15965 94490 | 64.6 19.9 15.5 5.5 40.9 61.9 | -26.104 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch4_dev_utt/ascore_12/dev_hires_o4.ctm.filt.sys
     %WER 40.7 | 13961 94495 | 64.4 18.8 16.8 5.0 40.7 70.4 | -26.622 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch4_dev_utt_offline/ascore_13/dev_hires_o4.ctm.filt.sys
 
-  #eval 
-    # epoch 0 
+  #eval
+    # epoch 0
     %WER 45.4 | 13992 89799 | 60.1 26.4 13.5 5.5 45.4 67.3 | -23.969 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch0_eval_utt/ascore_12/eval_hires_o4.ctm.filt.sys
     %WER 45.1 | 13893 89804 | 60.3 25.9 13.9 5.4 45.1 67.9 | -24.110 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch0_eval_utt_offline/ascore_12/eval_hires_o4.ctm.filt.sys
     # epoch 1
@@ -74,3 +75,56 @@ for x in exp/$mic/nnet2_online/*/decode*; do [ -d $x ] && [[ $x =~ "$1" ]] && gr
     %WER 44.2 | 13577 89767 | 61.1 22.3 16.6 5.3 44.2 68.9 | -25.003 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch4_eval_utt/ascore_11/eval_hires_o4.ctm.filt.sys
     %WER 44.0 | 13448 89769 | 60.8 21.4 17.8 4.9 44.0 69.6 | -25.331 | exp/mdm8/nnet2_online/nnet_ms_sp_smbr_0.000005/decode_epoch4_eval_utt_offline/ascore_12/eval_hires_o4.ctm.filt.sys
 
+
+#------------------------------------------------------------------------------------------------------------------------------------
+# Nnet3 systems
+
+# BLSTM + clean alignments
+#  local/nnet3/run_blstm.sh --mic mdm8\
+#    --chunk-right-context 20 \
+#    --use-sat-alignments true \
+#    --use-ihm-ali true
+# Note: Chunk right context of 20 limits the latency of the acoustic model to
+# 20 frames.
+
+
+  %WER 35.5 | 15221 94509 | 69.9 21.0 9.1 5.4 35.5 61.4 | -26.440 | exp/mdm8_cleanali/nnet3/lstm_bidirectional_ld0/decode_dev/ascore_10/dev_hires_o4.ctm.filt.sys
+  %WER 38.3 | 13423 89786 | 65.8 22.0 12.2 4.1 38.3 66.3 | -26.016 | exp/mdm8_cleanali/nnet3/lstm_bidirectional_ld0/decode_eval/ascore_10/eval_hires_o4.ctm.filt.sys
+
+
+################################# Chain Systems ######################
+
+  # local/chain/run_tdnn_ami_5.sh  --mic mdm8 --affix msl1.5_45wer
+   %WER 38.5 | 14761 94496 | 65.5 17.6 16.9 4.0 38.5 66.5 | 0.620 | exp/mdm8/chain/tdnn_ami5_msl1.5_45wer_sp/decode_dev/ascore_9/dev_hires_o4.ctm.filt.sys
+   %WER 41.5 | 14219 89974 | 62.2 18.5 19.2 3.7 41.5 65.6 | 0.596 | exp/mdm8/chain/tdnn_ami5_msl1.5_45wer_sp/decode_eval/ascore_9/eval_hires_o4.ctm.filt.sys
+
+  # local/chain/run_tdnn_ami_5.sh  --mic mdm8 --use-ihm-ali true --affix msl1.5_45wer
+   %WER 38.1 | 15296 94487 | 65.7 17.9 16.4 3.8 38.1 62.5 | 0.617 | exp/mdm8_cleanali/chain/tdnn_ami5_msl1.5_45wer_sp/decode_dev/ascore_9/dev_hires_o4.ctm.filt.sys
+   %WER 41.5 | 13795 89975 | 62.6 20.4 17.0 4.1 41.5 66.9 | 0.628 | exp/mdm8_cleanali/chain/tdnn_ami5_msl1.5_45wer_sp/decode_eval/ascore_8/eval_hires_o4.ctm.filt.sys
+
+#--------------------------------------------------------------------------------------------------------------------------------------------
+# Karel, JSALT 2015, (31.7.2015)
+# nnet1, MFCC-LDA-MLLT-DNN system (local/nnet/run_dnn_lda_mllt.sh),
+
+# dev,
+## GMM,
+%WER 59.1 | 14105 94500 | 47.5 34.3 18.2 6.6 59.1 76.0 | -22.348 | exp/mdm8/tri2a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_13/dev_o4.ctm.filt.sys
+%WER 57.2 | 14807 94503 | 49.6 33.2 17.3 6.8 57.2 72.1 | -22.450 | exp/mdm8/tri3a/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_13/dev_o4.ctm.filt.sys
+%WER 55.0 | 14511 94490 | 51.1 30.0 18.8 6.2 55.0 73.0 | -22.760 | exp/mdm8/tri3a_mmi_b0.1/decode_dev_2.mdl_ami_fsh.o3g.kn.pr1-7/ascore_11/dev_o4.ctm.filt.sys
+## DNN-Xent,
+%WER 48.2 | 15246 94513 | 58.4 28.7 12.9 6.7 48.2 67.3 | -23.329 | exp/mdm8/dnn4noSAT_pretrain-dbn_dnn/decode_dev_ami_fsh.o3g.kn.pr1-7/ascore_13/dev_o4.ctm.filt.sys
+## DNN-sMBR,
+%WER 46.2 | 15260 94500 | 60.1 26.0 13.9 6.3 46.2 66.3 | -23.908 | exp/mdm8/dnn4noSAT_pretrain-dbn_dnn_smbr/decode_dev_ami_fsh.o3g.kn.pr1-7_it1/ascore_13/dev_o4.ctm.filt.sys
+%WER 45.1 | 14204 94504 | 61.1 24.9 14.0 6.2 45.1 70.7 | -24.225 | exp/mdm8/dnn4noSAT_pretrain-dbn_dnn_smbr/decode_dev_ami_fsh.o3g.kn.pr1-7_it4/ascore_13/dev_o4.ctm.filt.sys
+
+# eval,
+## GMM,
+%WER 64.4 | 14362 90002 | 41.7 36.6 21.8 6.0 64.4 71.2 | -22.256 | exp/mdm8/tri2a/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_12/eval_o4.ctm.filt.sys
+%WER 62.1 | 13700 89987 | 44.0 35.5 20.5 6.2 62.1 74.1 | -22.267 | exp/mdm8/tri3a/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_12/eval_o4.ctm.filt.sys
+%WER 59.5 | 13822 89978 | 46.3 32.8 20.9 5.7 59.5 72.6 | -22.394 | exp/mdm8/tri3a_mmi_b0.1/decode_eval_2.mdl_ami_fsh.o3g.kn.pr1-7/ascore_10/eval_o4.ctm.filt.sys
+## DNN-Xent,
+%WER 52.1 | 13642 89829 | 53.6 30.7 15.7 5.7 52.1 71.6 | -22.884 | exp/mdm8/dnn4noSAT_pretrain-dbn_dnn/decode_eval_ami_fsh.o3g.kn.pr1-7/ascore_12/eval_o4.ctm.filt.sys
+## DNN-sMBR,
+%WER 50.3 | 14264 89966 | 54.7 27.6 17.8 5.0 50.3 67.5 | -23.397 | exp/mdm8/dnn4noSAT_pretrain-dbn_dnn_smbr/decode_eval_ami_fsh.o3g.kn.pr1-7_it1/ascore_13/eval_o4.ctm.filt.sys
+%WER 49.1 | 13969 89982 | 55.8 26.7 17.4 4.9 49.1 68.4 | -23.629 | exp/mdm8/dnn4noSAT_pretrain-dbn_dnn_smbr/decode_eval_ami_fsh.o3g.kn.pr1-7_it4/ascore_13/eval_o4.ctm.filt.sys
+