Hi,
I am trying to execute livepretend with a semicontinuous, context independent, 4 streams acoustic model, and it produce a fatal error below. Anyone has any clue, what is wrong? It seems like memory allocation problem... Thank you very much.
Hi,
Thank you very much for your many replies :D. I believe it should be correct since it is the default feature type provided when I initialized the codebook. Looks like manual debugging is unavoidable :(. Thanks again.
Tien Ping
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2005-03-02
The FATAL_ERROR line tells us that on line 773 of feat.c, it tried to allocate a buffer of -256 4-byte units. So look in that part of the code to understand why it calculated a negative buffer length. It obviously has something to do with the front end, but those parameters look OK to me.
I don't know much about Sphinx3, but are you sure that is supports use of a semicontinuous 4-stream acoustic model?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I am trying to execute livepretend with a semicontinuous, context independent, 4 streams acoustic model, and it produce a fatal error below. Anyone has any clue, what is wrong? It seems like memory allocation problem... Thank you very much.
-lextreedump 0 0
-lm /home/tan/sphinx3/sphinx3-0.5/model/lm/braf100/braf100.uni.arpa.DMP
-lmctlfn
-lmdumpdir
-lminmemory 0 0
-log3table 1 1
-logbase 1.0003 1.000300e+00
-lowerf 200 1.333333e+02
-lw 8.5 9.500000e+00
-machine_endian 0 0
-maxcdsenpf 100000 100000
-maxcepvecs 256 256
-maxhistpf 100 100
-maxhmmpf 20000 1500
-maxhyplen 1000 1000
-maxwpf 20 1
-mdef /home/tan/sphinx3/SphinxTrain/data/braf100_semi/braf100.mdef
-mean /home/tan/sphinx3/SphinxTrain/data/braf100_semi/braf100_s.ci_mean_6
-mixw /home/tan/sphinx3/SphinxTrain/data/braf100_semi/braf100_s.ci_mix_weight_6
-mixwfloor 0.0000001 1.000000e-07
-ncep 13 13
-nfft 256 512
-nfilt 31 40
-Nlextree 3 3
-outlatdir
-outlatoldfmt 1 1
-pbeam 1.0e-50 1.000000e-30
-pheurtype 0 0
-phypdump 1 1
-pl_beam 1.0e-80 1.000000e-80
-pl_window 1 1
-ptranskip 0 0
-rawext .raw .raw
-samprate 8000 16000
-senmgau .cont. .semi.
-silprob 0.1 1.000000e-01
-subvq
-subvqbeam 3.0e-3 1.000000e-02
-svq4svq 0 0
-tmat /home/tan/sphinx3/SphinxTrain/data/braf100_semi/braf100_s.ci_trans_matrix_6
-tmatfloor 0.0001 1.000000e-04
-treeugprob 1 1
-upperf 3500 6.855498e+03
-utt
-uw 0.7 7.000000e-01
-var /home/tan/sphinx3/SphinxTrain/data/braf100_semi/braf100_s.ci_variance_6
-varfloor 0.0001 1.000000e-04
-varnorm no no
-vqeval 3 3
-wbeam 1.0e-35 1.000000e-20
-wend_beam 1.0e-80 1.000000e-01
-wip 0.7 7.000000e-01
-wlen 0.0256 2.562500e-02
INFO: kbcore.c(95): Initializing core models:
INFO: logs3.c(102): Initializing logbase: 1.000300e+00 (add table: 1)
INFO: logs3.c(164): Log-Add table size = 29350
INFO: feat.c(642): Initializing feature stream to type: '4s_12c_24d_3p_12dd', CMN='current', VARNORM='no', AGC='max'
FATAL_ERROR: "ckd_alloc.c", line 95: calloc(-256,4) failed from feat.c(773)
TP
Hi,
Thank you very much for your many replies :D. I believe it should be correct since it is the default feature type provided when I initialized the codebook. Looks like manual debugging is unavoidable :(. Thanks again.
Tien Ping
The FATAL_ERROR line tells us that on line 773 of feat.c, it tried to allocate a buffer of -256 4-byte units. So look in that part of the code to understand why it calculated a negative buffer length. It obviously has something to do with the front end, but those parameters look OK to me.
I don't know much about Sphinx3, but are you sure that is supports use of a semicontinuous 4-stream acoustic model?
My mistake. I finally found the statement which sound something like "the s3.X decoder is only capable of handling continuous acoustic models".
Tien Ping