Hi All,
I want to train 31 languages supported in the dsmt tool and integrate the same in the imxrt106s Sln-local2-iot sdk.My doubt is can we change the MAX_INSTALLED_LANGUAGES from 4 to 31.Is there any issue in doing so?
When I added 5th language i am getting the following error,
region `SRAM_OC_CACHEABLE' overflowed by 24576 bytes
Following are the changes I have made for adding 5th language
1.Generated oob_demo folder in the local_voice folder of sln-local2-iot sdk using dsmt tool.
2.Added following lines in the source/sln_local_voice_model.s
oob_demo_ge_begin:
.incbin "./local_voice/oob_demo_ge/oob_demo_ge_pack.bin"
oob_demo_ge_end:
3.Added "extern unsigned int oob_demo_ge_begin;" in source/sln_local_voice_model.h
4.Added following lines in initialize_asr() function in source/sln_local_voice.c
langShell = appAsrShellCommands.multilingual;
langShell = appAsrShellCommands.multilingual;
#if MULTILINGUAL
// install multilingual
install_language(&g_asrControl, &g_asrLangModel[4], lang[4],
(unsigned char*) &oob_demo_ge_begin, 4);
#if MULTILINGUAL
install_inference_engine(&g_asrControl, &g_asrInfWW[4], lang[4], ASR_WW,
ww_ge, &g_memPoolWLang5[0],
WAKE_WORD_MEMPOOL_SIZE); // ww language5
SDK_ALIGN(
uint8_t __attribute__((section(".bss.$SRAM_OC_CACHEABLE"))) g_memPoolWLang5[WAKE_WORD_MEMPOOL_SIZE],
8);
My use case is add 31 language support and select one language from 31 during run time.
Kindly help me.
Thanks in advance
