int
main(int argc, char *argv[])
{
logmath_t *lmath;
cmd_ln_t *config;
acmod_t *acmod;
ps_mgau_t *ps;
ptm_mgau_t *s;
int i, lastcb;
lmath = logmath_init(1.0001, 0, 0);
config = cmd_ln_init(NULL, ps_args(), TRUE,
"-compallsen", "yes",
"-input_endian", "little",
NULL);
cmd_ln_parse_file_r(config, ps_args(), MODELDIR "/en-us/en-us/feat.params", FALSE);
cmd_ln_set_str_extra_r(config, "_mdef", MODELDIR "/en-us/en-us/mdef");
cmd_ln_set_str_extra_r(config, "_mean", MODELDIR "/en-us/en-us/means");
cmd_ln_set_str_extra_r(config, "_var", MODELDIR "/en-us/en-us/variances");
cmd_ln_set_str_extra_r(config, "_tmat", MODELDIR "/en-us/en-us/transition_matrices");
cmd_ln_set_str_extra_r(config, "_sendump", MODELDIR "/en-us/en-us/sendump");
cmd_ln_set_str_extra_r(config, "_mixw", NULL);
cmd_ln_set_str_extra_r(config, "_lda", NULL);
cmd_ln_set_str_extra_r(config, "_senmgau", NULL);
err_set_debug_level(3);
TEST_ASSERT(config);
TEST_ASSERT((acmod = acmod_init(config, lmath, NULL, NULL)));
TEST_ASSERT((ps = acmod->mgau));
TEST_EQUAL(0, strcmp(ps->vt->name, "ptm"));
s = (ptm_mgau_t *)ps;
E_DEBUG(2,("PTM model loaded: %d codebooks, %d senones, %d frames of history\n",
s->g->n_mgau, s->n_sen, s->n_fast_hist));
E_DEBUG(2,("Senone to codebook mappings:\n"));
lastcb = s->sen2cb[0];
E_DEBUG(2,("\t%d: 0", lastcb));
for (i = 0; i < s->n_sen; ++i) {
if (s->sen2cb[i] != lastcb) {
lastcb = s->sen2cb[i];
E_DEBUGCONT(2,("-%d\n", i-1));
E_DEBUGCONT(2,("\t%d: %d", lastcb, i));
}
}
E_INFOCONT("-%d\n", i-1);
run_acmod_test(acmod);
#if 0
/* Replace it with ms_mgau. */
ptm_mgau_free(ps);
cmd_ln_set_str_r(config,
"-mixw",
MODELDIR "/en-us/en-us/mixture_weights");
TEST_ASSERT((acmod->mgau = ms_mgau_init(acmod, lmath, acmod->mdef)));
run_acmod_test(acmod);
cmd_ln_free_r(config);
#endif
return 0;
}
int
main(int argc, char *argv[])
{
cmd_ln_t *config;
ngram_model_t *lm = NULL;
logmath_t *lmath;
int itype, otype;
char const *kase;
if ((config = cmd_ln_parse_r(NULL, defn, argc, argv, TRUE)) == NULL)
return 1;
if (cmd_ln_boolean_r(config, "-help")) {
usagemsg(argv[0]);
}
err_set_debug_level(cmd_ln_int32_r(config, "-debug"));
/* Create log math object. */
if ((lmath = logmath_init
(cmd_ln_float64_r(config, "-logbase"), 0, 0)) == NULL) {
E_FATAL("Failed to initialize log math\n");
}
if (cmd_ln_str_r(config, "-i") == NULL || cmd_ln_str_r(config, "-i") == NULL) {
E_ERROR("Please specify both input and output models\n");
goto error_out;
}
/* Load the input language model. */
if (cmd_ln_str_r(config, "-ifmt")) {
if ((itype = ngram_str_to_type(cmd_ln_str_r(config, "-ifmt")))
== NGRAM_INVALID) {
E_ERROR("Invalid input type %s\n", cmd_ln_str_r(config, "-ifmt"));
goto error_out;
}
lm = ngram_model_read(config, cmd_ln_str_r(config, "-i"),
itype, lmath);
}
else {
lm = ngram_model_read(config, cmd_ln_str_r(config, "-i"),
NGRAM_AUTO, lmath);
}
/* Guess or set the output language model type. */
if (cmd_ln_str_r(config, "-ofmt")) {
if ((otype = ngram_str_to_type(cmd_ln_str_r(config, "-ofmt")))
== NGRAM_INVALID) {
E_ERROR("Invalid output type %s\n", cmd_ln_str_r(config, "-ofmt"));
goto error_out;
}
}
else {
otype = ngram_file_name_to_type(cmd_ln_str_r(config, "-o"));
}
/* Recode the language model if desired. */
if (cmd_ln_str_r(config, "-ienc")) {
if (ngram_model_recode(lm, cmd_ln_str_r(config, "-ienc"),
cmd_ln_str_r(config, "-oenc")) != 0) {
E_ERROR("Failed to recode language model from %s to %s\n",
cmd_ln_str_r(config, "-ienc"),
cmd_ln_str_r(config, "-oenc"));
goto error_out;
}
}
/* Case fold if requested. */
if ((kase = cmd_ln_str_r(config, "-case"))) {
if (0 == strcmp(kase, "lower")) {
ngram_model_casefold(lm, NGRAM_LOWER);
}
else if (0 == strcmp(kase, "upper")) {
ngram_model_casefold(lm, NGRAM_UPPER);
}
else {
E_ERROR("Unknown value for -case: %s\n", kase);
goto error_out;
}
}
/* Write the output language model. */
if (ngram_model_write(lm, cmd_ln_str_r(config, "-o"), otype) != 0) {
E_ERROR("Failed to write language model in format %s to %s\n",
ngram_type_to_str(otype), cmd_ln_str_r(config, "-o"));
goto error_out;
}
/* That's all folks! */
ngram_model_free(lm);
return 0;
error_out:
ngram_model_free(lm);
return 1;
}
int
ps_reinit(ps_decoder_t *ps, cmd_ln_t *config)
{
const char *path;
const char *keyphrase;
int32 lw;
if (config && config != ps->config) {
cmd_ln_free_r(ps->config);
ps->config = cmd_ln_retain(config);
}
err_set_debug_level(cmd_ln_int32_r(ps->config, "-debug"));
ps->mfclogdir = cmd_ln_str_r(ps->config, "-mfclogdir");
ps->rawlogdir = cmd_ln_str_r(ps->config, "-rawlogdir");
ps->senlogdir = cmd_ln_str_r(ps->config, "-senlogdir");
/* Fill in some default arguments. */
ps_init_defaults(ps);
/* Free old searches (do this before other reinit) */
ps_free_searches(ps);
ps->searches = hash_table_new(3, HASH_CASE_YES);
/* Free old acmod. */
acmod_free(ps->acmod);
ps->acmod = NULL;
/* Free old dictionary (must be done after the two things above) */
dict_free(ps->dict);
ps->dict = NULL;
/* Free d2p */
dict2pid_free(ps->d2p);
ps->d2p = NULL;
/* Logmath computation (used in acmod and search) */
if (ps->lmath == NULL
|| (logmath_get_base(ps->lmath) !=
(float64)cmd_ln_float32_r(ps->config, "-logbase"))) {
if (ps->lmath)
logmath_free(ps->lmath);
ps->lmath = logmath_init
((float64)cmd_ln_float32_r(ps->config, "-logbase"), 0,
cmd_ln_boolean_r(ps->config, "-bestpath"));
}
/* Acoustic model (this is basically everything that
* uttproc.c, senscr.c, and others used to do) */
if ((ps->acmod = acmod_init(ps->config, ps->lmath, NULL, NULL)) == NULL)
return -1;
if (cmd_ln_int32_r(ps->config, "-pl_window") > 0) {
/* Initialize an auxiliary phone loop search, which will run in
* "parallel" with FSG or N-Gram search. */
if ((ps->phone_loop =
phone_loop_search_init(ps->config, ps->acmod, ps->dict)) == NULL)
return -1;
hash_table_enter(ps->searches,
ckd_salloc(ps_search_name(ps->phone_loop)),
ps->phone_loop);
}
/* Dictionary and triphone mappings (depends on acmod). */
/* FIXME: pass config, change arguments, implement LTS, etc. */
if ((ps->dict = dict_init(ps->config, ps->acmod->mdef, ps->acmod->lmath)) == NULL)
return -1;
if ((ps->d2p = dict2pid_build(ps->acmod->mdef, ps->dict)) == NULL)
return -1;
lw = cmd_ln_float32_r(config, "-lw");
/* Determine whether we are starting out in FSG or N-Gram search mode.
* If neither is used skip search initialization. */
/* Load KWS if one was specified in config */
if ((keyphrase = cmd_ln_str_r(config, "-keyphrase"))) {
if (ps_set_keyphrase(ps, PS_DEFAULT_SEARCH, keyphrase))
return -1;
ps_set_search(ps, PS_DEFAULT_SEARCH);
}
if ((path = cmd_ln_str_r(config, "-kws"))) {
if (ps_set_kws(ps, PS_DEFAULT_SEARCH, path))
return -1;
ps_set_search(ps, PS_DEFAULT_SEARCH);
}
/* Load an FSG if one was specified in config */
if ((path = cmd_ln_str_r(config, "-fsg"))) {
fsg_model_t *fsg = fsg_model_readfile(path, ps->lmath, lw);
if (!fsg)
return -1;
if (ps_set_fsg(ps, PS_DEFAULT_SEARCH, fsg))
return -1;
ps_set_search(ps, PS_DEFAULT_SEARCH);
}
/* Or load a JSGF grammar */
if ((path = cmd_ln_str_r(config, "-jsgf"))) {
if (ps_set_jsgf_file(ps, PS_DEFAULT_SEARCH, path)
|| ps_set_search(ps, PS_DEFAULT_SEARCH))
return -1;
}
if ((path = cmd_ln_str_r(ps->config, "-allphone"))) {
if (ps_set_allphone_file(ps, PS_DEFAULT_SEARCH, path)
|| ps_set_search(ps, PS_DEFAULT_SEARCH))
return -1;
}
if ((path = cmd_ln_str_r(ps->config, "-lm")) &&
!cmd_ln_boolean_r(ps->config, "-allphone")) {
if (ps_set_lm_file(ps, PS_DEFAULT_SEARCH, path)
|| ps_set_search(ps, PS_DEFAULT_SEARCH))
return -1;
}
if ((path = cmd_ln_str_r(ps->config, "-lmctl"))) {
const char *name;
ngram_model_t *lmset;
ngram_model_set_iter_t *lmset_it;
if (!(lmset = ngram_model_set_read(ps->config, path, ps->lmath))) {
E_ERROR("Failed to read language model control file: %s\n", path);
return -1;
}
for(lmset_it = ngram_model_set_iter(lmset);
lmset_it; lmset_it = ngram_model_set_iter_next(lmset_it)) {
ngram_model_t *lm = ngram_model_set_iter_model(lmset_it, &name);
E_INFO("adding search %s\n", name);
if (ps_set_lm(ps, name, lm)) {
ngram_model_free(lm);
ngram_model_set_iter_free(lmset_it);
return -1;
}
ngram_model_free(lm);
}
name = cmd_ln_str_r(config, "-lmname");
if (name)
ps_set_search(ps, name);
else {
E_ERROR("No default LM name (-lmname) for `-lmctl'\n");
return -1;
}
}
/* Initialize performance timer. */
ps->perf.name = "decode";
ptmr_init(&ps->perf);
return 0;
}
void
kb_init(kb_t * kb, cmd_ln_t *config)
{
kbcore_t *kbcore;
mdef_t *mdef;
dict_t *dict;
dict2pid_t *d2p;
int32 cisencnt;
/* STRUCTURE: Initialize the kb structure to zero, just in case */
memset(kb, 0, sizeof(*kb));
kb->kbcore = kbcore_init(config);
if (kb->kbcore == NULL)
E_FATAL("Initialization of kb failed\n");
kbcore = kb->kbcore;
mdef = kbcore_mdef(kbcore);
dict = kbcore_dict(kbcore);
d2p = kbcore_dict2pid(kbcore);
err_set_debug_level(cmd_ln_int32_r(config, "-debug"));
/* STRUCTURE INITIALIZATION: Initialize the beam data structure */
if (cmd_ln_exists_r(config, "-ptranskip")) {
kb->beam = beam_init(cmd_ln_float64_r(config, "-beam"),
cmd_ln_float64_r(config, "-pbeam"),
cmd_ln_float64_r(config, "-wbeam"),
cmd_ln_float64_r(config, "-wend_beam"),
cmd_ln_int32_r(config, "-ptranskip"), mdef_n_ciphone(mdef),
kbcore->logmath
);
/* REPORT : Report the parameters in the beam data structure */
if (REPORT_KB)
beam_report(kb->beam);
}
/* STRUCTURE INITIALIZATION: Initialize the fast GMM computation data structure */
if (cmd_ln_exists_r(config, "-ci_pbeam")) {
kb->fastgmm = fast_gmm_init(cmd_ln_int32_r(config, "-ds"),
cmd_ln_int32_r(config, "-cond_ds"),
cmd_ln_int32_r(config, "-dist_ds"),
cmd_ln_int32_r(config, "-gs4gs"),
cmd_ln_int32_r(config, "-svq4svq"),
cmd_ln_float64_r(config, "-subvqbeam"),
cmd_ln_float64_r(config, "-ci_pbeam"),
cmd_ln_float64_r(config, "-tighten_factor"),
cmd_ln_int32_r(config, "-maxcdsenpf"),
mdef->n_ci_sen,
kbcore->logmath);
/* REPORT : Report the parameters in the fast_gmm_t data struture */
if (REPORT_KB)
fast_gmm_report(kb->fastgmm);
}
/* STRUCTURE INITIALIZATION: Initialize the phoneme lookahead data structure */
if (cmd_ln_exists_r(config, "-pl_beam")) {
kb->pl = pl_init(cmd_ln_int32_r(config, "-pheurtype"),
cmd_ln_float64_r(config, "-pl_beam"), mdef_n_ciphone(mdef),
kbcore->logmath
);
/* REPORT : Report the parameters in the pl_t data struture */
if (REPORT_KB)
pl_report(kb->pl);
}
/* STRUCTURE INITIALIZATION: Initialize the acoustic score data structure */
{
int32 pl_window = 1;
if (cmd_ln_exists_r(config, "-pl_window"))
pl_window = cmd_ln_int32_r(config, "-pl_window");
for (cisencnt = 0; cisencnt == mdef->cd2cisen[cisencnt]; cisencnt++) ;
kb->ascr = ascr_init(kbcore_n_mgau(kbcore),
kb->kbcore->dict2pid->n_comstate,
mdef_n_sseq(mdef),
dict2pid_n_comsseq(d2p),
pl_window, cisencnt);
if (REPORT_KB)
ascr_report(kb->ascr);
}
/* Initialize the front end if -adcin is specified */
if (cmd_ln_exists_r(config, "-adcin") && cmd_ln_boolean_r(config, "-adcin")) {
if ((kb->fe = fe_init_auto_r(config)) == NULL) {
E_FATAL("fe_init_auto_r() failed\n");
}
}
/* STRUCTURE INITIALIZATION : The feature vector */
if ((kb->feat =
feat_array_alloc(kbcore_fcb(kbcore), S3_MAX_FRAMES)) == NULL)
E_FATAL("feat_array_alloc() failed\n");
/* STRUCTURE INITIALIZATION : The statistics for the search */
kb->stat = stat_init();
/* STRUCTURE INITIALIZATION : The adaptation routines of the search */
kb->adapt_am = adapt_am_init();
if (cmd_ln_str_r(config, "-mllr")) {
kb_setmllr(cmd_ln_str_r(config, "-mllr"), cmd_ln_str_r(config, "-cb2mllr"), kb);
}
/* CHECK: make sure when (-cond_ds) is specified, a Gaussian map is also specified */
if (cmd_ln_int32_r(config, "-cond_ds") > 0 && kb->kbcore->gs == NULL)
E_FATAL
("Conditional Down Sampling require the use of Gaussian Selection map\n");
/* MEMORY ALLOCATION : Word best score and exit */
/* Open hypseg file if specified */
kb->matchsegfp = kb->matchfp = NULL;
kb->matchsegfp = file_open(cmd_ln_str_r(config, "-hypseg"));
kb->matchfp = file_open(cmd_ln_str_r(config, "-hyp"));
if (cmd_ln_exists_r(config, "-hmmdump"))
kb->hmmdumpfp = cmd_ln_int32_r(config, "-hmmdump") ? stderr : NULL;
/* STRUCTURE INITIALIZATION : The search data structure, done only
after kb is initialized kb is acted as a clipboard. */
if (cmd_ln_exists_r(config, "-op_mode")) {
/* -op_mode, if set (i.e. not -1), takes precedence over -mode. */
if (cmd_ln_int32_r(config, "-op_mode") != -1)
kb->op_mode = cmd_ln_int32_r(config, "-op_mode");
else
kb->op_mode = srch_mode_str_to_index(cmd_ln_str_r(config, "-mode"));
E_INFO("SEARCH MODE INDEX %d\n", kb->op_mode);
if ((kb->srch = (srch_t *) srch_init(kb, kb->op_mode)) == NULL) {
E_FATAL("Search initialization failed. Forced exit\n");
}
if (REPORT_KB) {
srch_report(kb->srch);
}
}
}
int
ps_reinit(ps_decoder_t *ps, cmd_ln_t *config)
{
char const *lmfile, *lmctl = NULL;
if (config && config != ps->config) {
cmd_ln_free_r(ps->config);
ps->config = config;
}
#ifndef _WIN32_WCE
/* Set up logging. */
if (cmd_ln_str_r(ps->config, "-logfn"))
err_set_logfile(cmd_ln_str_r(ps->config, "-logfn"));
#endif
err_set_debug_level(cmd_ln_int32_r(ps->config, "-debug"));
ps->mfclogdir = cmd_ln_str_r(ps->config, "-mfclogdir");
ps->rawlogdir = cmd_ln_str_r(ps->config, "-rawlogdir");
/* Fill in some default arguments. */
ps_init_defaults(ps);
/* Free old searches (do this before other reinit) */
ps_free_searches(ps);
/* Free old acmod. */
acmod_free(ps->acmod);
ps->acmod = NULL;
/* Free old dictionary (must be done after the two things above) */
dict_free(ps->dict);
ps->dict = NULL;
/* Logmath computation (used in acmod and search) */
if (ps->lmath == NULL
|| (logmath_get_base(ps->lmath) !=
(float64)cmd_ln_float32_r(ps->config, "-logbase"))) {
if (ps->lmath)
logmath_free(ps->lmath);
ps->lmath = logmath_init
((float64)cmd_ln_float32_r(ps->config, "-logbase"), 0,
cmd_ln_boolean_r(ps->config, "-bestpath"));
}
/* Acoustic model (this is basically everything that
* uttproc.c, senscr.c, and others used to do) */
if ((ps->acmod = acmod_init(ps->config, ps->lmath, NULL, NULL)) == NULL)
return -1;
/* Make the acmod's feature buffer growable if we are doing two-pass search. */
if (cmd_ln_boolean_r(ps->config, "-fwdflat")
&& cmd_ln_boolean_r(ps->config, "-fwdtree"))
acmod_set_grow(ps->acmod, TRUE);
if ((ps->pl_window = cmd_ln_int32_r(ps->config, "-pl_window"))) {
/* Initialize an auxiliary phone loop search, which will run in
* "parallel" with FSG or N-Gram search. */
if ((ps->phone_loop = phone_loop_search_init(ps->config,
ps->acmod, ps->dict)) == NULL)
return -1;
ps->searches = glist_add_ptr(ps->searches, ps->phone_loop);
}
/* Dictionary and triphone mappings (depends on acmod). */
/* FIXME: pass config, change arguments, implement LTS, etc. */
if ((ps->dict = dict_init(ps->config, ps->acmod->mdef)) == NULL)
return -1;
/* Determine whether we are starting out in FSG or N-Gram search mode. */
if (cmd_ln_str_r(ps->config, "-fsg") || cmd_ln_str_r(ps->config, "-jsgf")) {
ps_search_t *fsgs;
if ((ps->d2p = dict2pid_build(ps->acmod->mdef, ps->dict)) == NULL)
return -1;
if ((fsgs = fsg_search_init(ps->config, ps->acmod, ps->dict, ps->d2p)) == NULL)
return -1;
fsgs->pls = ps->phone_loop;
ps->searches = glist_add_ptr(ps->searches, fsgs);
ps->search = fsgs;
}
else if ((lmfile = cmd_ln_str_r(ps->config, "-lm"))
|| (lmctl = cmd_ln_str_r(ps->config, "-lmctl"))) {
ps_search_t *ngs;
if ((ps->d2p = dict2pid_build(ps->acmod->mdef, ps->dict)) == NULL)
return -1;
if ((ngs = ngram_search_init(ps->config, ps->acmod, ps->dict, ps->d2p)) == NULL)
return -1;
ngs->pls = ps->phone_loop;
ps->searches = glist_add_ptr(ps->searches, ngs);
ps->search = ngs;
}
/* Otherwise, we will initialize the search whenever the user
* decides to load an FSG or a language model. */
else {
if ((ps->d2p = dict2pid_build(ps->acmod->mdef, ps->dict)) == NULL)
return -1;
}
/* Initialize performance timer. */
ps->perf.name = "decode";
ptmr_init(&ps->perf);
return 0;
}
