Hi –
The Trition Inference server that is packaged as part of the Riva Quickstart Container is BSD licensed. The riva_server
binary that comes packaged as part of the container (nvcr.io/nvidia/riva/riva-speech:1.6.0-beta-server
) used with Riva Quickstart does not have any source code available (that I am aware of). Is there any intention of releasing that source code under an open source license or making it generally available?
The riva-speech
container additionally has a number of compiled backends for Trition:
ls -l backends/
total 48252
-r-xr-xr-x 1 riva-server riva-server 2536024 Sep 28 02:11 libtriton_riva_asr_decoder_cpu.so
-r-xr-xr-x 1 riva-server riva-server 5990432 Sep 28 02:11 libtriton_riva_asr_decoder_gpu.so
-r-xr-xr-x 1 riva-server riva-server 594720 Sep 28 02:11 libtriton_riva_asr_features.so
-r-xr-xr-x 1 riva-server riva-server 3503192 Sep 28 02:11 libtriton_riva_asr_lattices.so
-r-xr-xr-x 1 riva-server riva-server 506136 Sep 28 02:11 libtriton_riva_asr_vad.so
-r-xr-xr-x 1 riva-server riva-server 2366312 Sep 28 02:11 libtriton_riva_nlp_detokenizer.so
-r-xr-xr-x 1 riva-server riva-server 2356784 Sep 28 02:11 libtriton_riva_nlp_labels.so
-r-xr-xr-x 1 riva-server riva-server 2370416 Sep 28 02:11 libtriton_riva_nlp_punctuation.so
-r-xr-xr-x 1 riva-server riva-server 419776 Sep 28 02:11 libtriton_riva_nlp_punctuation_postprocessor.so
-r-xr-xr-x 1 riva-server riva-server 6984304 Sep 28 02:11 libtriton_riva_nlp_qa.so
-r-xr-xr-x 1 riva-server riva-server 2356160 Sep 28 02:11 libtriton_riva_nlp_seqlabel.so
-r-xr-xr-x 1 riva-server riva-server 5684488 Sep 28 02:11 libtriton_riva_nlp_tokenizer.so
-r-xr-xr-x 1 riva-server riva-server 309832 Sep 28 02:11 libtriton_riva_tts_chunker.so
-r-xr-xr-x 1 riva-server riva-server 1080008 Sep 28 02:11 libtriton_riva_tts_postprocessor.so
-r-xr-xr-x 1 riva-server riva-server 5400568 Sep 28 02:11 libtriton_riva_tts_preprocessor.so
-r-xr-xr-x 1 riva-server riva-server 6915944 Sep 28 02:11 libtriton_riva_tts_taco_postnet.so
Will any of this code become open source or available?
Thank you!