csukuangfj
commited on
Commit
•
8fef599
1
Parent(s):
355b986
update model
Browse files- sherpa-onnx-asr.js +71 -4
sherpa-onnx-asr.js
CHANGED
@@ -35,6 +35,10 @@ function freeConfig(config, Module) {
|
|
35 |
freeConfig(config.whisper, Module)
|
36 |
}
|
37 |
|
|
|
|
|
|
|
|
|
38 |
if ('tdnn' in config) {
|
39 |
freeConfig(config.tdnn, Module)
|
40 |
}
|
@@ -563,7 +567,7 @@ function initSherpaOnnxOfflineWhisperModelConfig(config, Module) {
|
|
563 |
const n = encoderLen + decoderLen + languageLen + taskLen;
|
564 |
const buffer = Module._malloc(n);
|
565 |
|
566 |
-
const len = 5 * 4; // 4 pointers
|
567 |
const ptr = Module._malloc(len);
|
568 |
|
569 |
let offset = 0;
|
@@ -598,6 +602,55 @@ function initSherpaOnnxOfflineWhisperModelConfig(config, Module) {
|
|
598 |
}
|
599 |
}
|
600 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
601 |
function initSherpaOnnxOfflineTdnnModelConfig(config, Module) {
|
602 |
const n = Module.lengthBytesUTF8(config.model || '') + 1;
|
603 |
const buffer = Module._malloc(n);
|
@@ -693,6 +746,15 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
|
|
693 |
};
|
694 |
}
|
695 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
696 |
if (!('tdnn' in config)) {
|
697 |
config.tdnn = {
|
698 |
model: '',
|
@@ -724,8 +786,11 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
|
|
724 |
const senseVoice =
|
725 |
initSherpaOnnxOfflineSenseVoiceModelConfig(config.senseVoice, Module);
|
726 |
|
|
|
|
|
|
|
727 |
const len = transducer.len + paraformer.len + nemoCtc.len + whisper.len +
|
728 |
-
tdnn.len + 8 * 4 + senseVoice.len;
|
729 |
|
730 |
const ptr = Module._malloc(len);
|
731 |
|
@@ -745,7 +810,6 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
|
|
745 |
Module._CopyHeap(tdnn.ptr, tdnn.len, ptr + offset);
|
746 |
offset += tdnn.len;
|
747 |
|
748 |
-
|
749 |
const tokensLen = Module.lengthBytesUTF8(config.tokens || '') + 1;
|
750 |
const providerLen = Module.lengthBytesUTF8(config.provider || 'cpu') + 1;
|
751 |
const modelTypeLen = Module.lengthBytesUTF8(config.modelType || '') + 1;
|
@@ -817,11 +881,14 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
|
|
817 |
offset += 4;
|
818 |
|
819 |
Module._CopyHeap(senseVoice.ptr, senseVoice.len, ptr + offset);
|
|
|
|
|
|
|
820 |
|
821 |
return {
|
822 |
buffer: buffer, ptr: ptr, len: len, transducer: transducer,
|
823 |
paraformer: paraformer, nemoCtc: nemoCtc, whisper: whisper, tdnn: tdnn,
|
824 |
-
senseVoice: senseVoice,
|
825 |
}
|
826 |
}
|
827 |
|
|
|
35 |
freeConfig(config.whisper, Module)
|
36 |
}
|
37 |
|
38 |
+
if ('moonshine' in config) {
|
39 |
+
freeConfig(config.moonshine, Module)
|
40 |
+
}
|
41 |
+
|
42 |
if ('tdnn' in config) {
|
43 |
freeConfig(config.tdnn, Module)
|
44 |
}
|
|
|
567 |
const n = encoderLen + decoderLen + languageLen + taskLen;
|
568 |
const buffer = Module._malloc(n);
|
569 |
|
570 |
+
const len = 5 * 4; // 4 pointers + 1 int32
|
571 |
const ptr = Module._malloc(len);
|
572 |
|
573 |
let offset = 0;
|
|
|
602 |
}
|
603 |
}
|
604 |
|
605 |
+
function initSherpaOnnxOfflineMoonshineModelConfig(config, Module) {
|
606 |
+
const preprocessorLen = Module.lengthBytesUTF8(config.preprocessor || '') + 1;
|
607 |
+
const encoderLen = Module.lengthBytesUTF8(config.encoder || '') + 1;
|
608 |
+
const uncachedDecoderLen =
|
609 |
+
Module.lengthBytesUTF8(config.uncachedDecoder || '') + 1;
|
610 |
+
const cachedDecoderLen =
|
611 |
+
Module.lengthBytesUTF8(config.cachedDecoder || '') + 1;
|
612 |
+
|
613 |
+
const n =
|
614 |
+
preprocessorLen + encoderLen + uncachedDecoderLen + cachedDecoderLen;
|
615 |
+
const buffer = Module._malloc(n);
|
616 |
+
|
617 |
+
const len = 4 * 4; // 4 pointers
|
618 |
+
const ptr = Module._malloc(len);
|
619 |
+
|
620 |
+
let offset = 0;
|
621 |
+
Module.stringToUTF8(
|
622 |
+
config.preprocessor || '', buffer + offset, preprocessorLen);
|
623 |
+
offset += preprocessorLen;
|
624 |
+
|
625 |
+
Module.stringToUTF8(config.encoder || '', buffer + offset, encoderLen);
|
626 |
+
offset += encoderLen;
|
627 |
+
|
628 |
+
Module.stringToUTF8(
|
629 |
+
config.uncachedDecoder || '', buffer + offset, uncachedDecoderLen);
|
630 |
+
offset += uncachedDecoderLen;
|
631 |
+
|
632 |
+
Module.stringToUTF8(
|
633 |
+
config.cachedDecoder || '', buffer + offset, cachedDecoderLen);
|
634 |
+
offset += cachedDecoderLen;
|
635 |
+
|
636 |
+
offset = 0;
|
637 |
+
Module.setValue(ptr, buffer + offset, 'i8*');
|
638 |
+
offset += preprocessorLen;
|
639 |
+
|
640 |
+
Module.setValue(ptr + 4, buffer + offset, 'i8*');
|
641 |
+
offset += encoderLen;
|
642 |
+
|
643 |
+
Module.setValue(ptr + 8, buffer + offset, 'i8*');
|
644 |
+
offset += uncachedDecoderLen;
|
645 |
+
|
646 |
+
Module.setValue(ptr + 12, buffer + offset, 'i8*');
|
647 |
+
offset += cachedDecoderLen;
|
648 |
+
|
649 |
+
return {
|
650 |
+
buffer: buffer, ptr: ptr, len: len,
|
651 |
+
}
|
652 |
+
}
|
653 |
+
|
654 |
function initSherpaOnnxOfflineTdnnModelConfig(config, Module) {
|
655 |
const n = Module.lengthBytesUTF8(config.model || '') + 1;
|
656 |
const buffer = Module._malloc(n);
|
|
|
746 |
};
|
747 |
}
|
748 |
|
749 |
+
if (!('moonshine' in config)) {
|
750 |
+
config.moonshine = {
|
751 |
+
preprocessor: '',
|
752 |
+
encoder: '',
|
753 |
+
uncachedDecoder: '',
|
754 |
+
cachedDecoder: '',
|
755 |
+
};
|
756 |
+
}
|
757 |
+
|
758 |
if (!('tdnn' in config)) {
|
759 |
config.tdnn = {
|
760 |
model: '',
|
|
|
786 |
const senseVoice =
|
787 |
initSherpaOnnxOfflineSenseVoiceModelConfig(config.senseVoice, Module);
|
788 |
|
789 |
+
const moonshine =
|
790 |
+
initSherpaOnnxOfflineMoonshineModelConfig(config.moonshine, Module);
|
791 |
+
|
792 |
const len = transducer.len + paraformer.len + nemoCtc.len + whisper.len +
|
793 |
+
tdnn.len + 8 * 4 + senseVoice.len + moonshine.len;
|
794 |
|
795 |
const ptr = Module._malloc(len);
|
796 |
|
|
|
810 |
Module._CopyHeap(tdnn.ptr, tdnn.len, ptr + offset);
|
811 |
offset += tdnn.len;
|
812 |
|
|
|
813 |
const tokensLen = Module.lengthBytesUTF8(config.tokens || '') + 1;
|
814 |
const providerLen = Module.lengthBytesUTF8(config.provider || 'cpu') + 1;
|
815 |
const modelTypeLen = Module.lengthBytesUTF8(config.modelType || '') + 1;
|
|
|
881 |
offset += 4;
|
882 |
|
883 |
Module._CopyHeap(senseVoice.ptr, senseVoice.len, ptr + offset);
|
884 |
+
offset += senseVoice.len;
|
885 |
+
|
886 |
+
Module._CopyHeap(moonshine.ptr, moonshine.len, ptr + offset);
|
887 |
|
888 |
return {
|
889 |
buffer: buffer, ptr: ptr, len: len, transducer: transducer,
|
890 |
paraformer: paraformer, nemoCtc: nemoCtc, whisper: whisper, tdnn: tdnn,
|
891 |
+
senseVoice: senseVoice, moonshine: moonshine,
|
892 |
}
|
893 |
}
|
894 |
|