Command-line Clients¶

A Docker image containing sample command-line drivers for the Jarvis services is available on NGC. Pull and run the container by using the commands below. The client expects that a Jarvis server is running with models deployed, and all command-line drivers accept an optional argument to specify the location of the server. No GPU is required to run the sample clients.

docker pull nvcr.io/nvidia/jarvis/jarvis-speech-client:1.0.0-b.2
docker run -it --rm nvcr.io/nvidia/jarvis/jarvis-speech-client:1.0.0-b.2

Speech Recognition¶

Both binary C++ and Python clients are included in the Docker image with options to do ASR inference in streaming as well as offline (non-streaming) mode.

C++ Streaming Example¶

Running the C++ ASR streaming client, where --audio_file is used to specify the audio file that is to be transcribed. Other options can be found in jarvis_streaming_asr_client.cc.

jarvis_streaming_asr_client --audio_file /work/wav/test/1272-135031-0001.wav

C++ Offline/Batch (non-streaming) Example¶

Running the C++ ASR offline client, where --audio_file is used to specify the audio file that is to be transcribed. Other options can be found in jarvis_asr_client.cc.

jarvis_asr_client --audio_file /work/wav/test/1272-135031-0001.wav

Python Streaming Example¶

Running the Python ASR streaming client, where --input-file is used to specify the audio file that is to be transcribed. Other options can be found in jarvis_streaming_asr_client.py.

python jarvis_streaming_asr_client.py --input-file=/work/wav/test/1272-135031-0001.wav

Other Python ASR clients are also available. The transcribe_mic.py example is the preferred, cross-platform way to interact with Jarvis using a microphone from an arbitrary remote system running on any operating system and hardware platform.

transcribe_file.py transcribes audio files in streaming mode.
transcribe_file_offline.py transcribes audio files in offline mode.
transcribe_file_rt.py transcribes audio files while playing back through PyAudio stream.
transcribe_mic.py transcribes audio from a microphone device.

Natural Language Processing¶

Both binary C++ and Python clients are included in the Docker image for supported NLP tasks.

C++ NER Example¶

Running C++ Token Classification Model (NER) where --queries is used to specify the query file in which to provide token labels for. Other options can be found in jarvis_nlp_classify_tokens.cc.

jarvis_nlp_classify_tokens --queries=/work/test_files/nlu/queries.txt

C++ QA Example¶

Running the C++ Question Answering client where --questions is used to specify the file containing questions and --contexts is used to specify the file containing the context paragraph. Other options can be found in jarvis_nlp_qa.cc.

jarvis_nlp_qa --questions=/work/test_files/nlu/qa_questions.txt --contexts=/work/test_files/nlu/qa_contexts.txt

Python QA Example¶

Running the Python Question Answering client. Modify test_context and req.query in the test_qa.py script to try different QA examples.

python test_qa.py

C++ Punctuation Example¶

Running the C++ Punctuation client where --queries is used to specify the file containing inputs to punctuate. Note that, if you want to specify --output to write output of the punctuation model to a file, you need to also specify --parallel_requests=1.

jarvis_nlp_punct --queries=/work/test_files/nlu/punctuation_input.txt

Speech Synthesis¶

Binary C++ clients are supported in the Docker image for TTS.

C++ TTS Client Example¶

Running the C++ TTS client where --text is used to specify the input text for which to synthesize audio. The output of the client is a wav file which can be specified by --audio_file. Other options can be found in jarvis_tts_client.cc.

jarvis_tts_client --text="I had a dream yesterday."

C++ TTS Performance Client Example¶

Running the C++ TTS Performance client which will provide information about latency and throughput. Options --text can be used to specify the input text and --text_file can be used to specify the file containing multiple text inputs. Other options can be found in jarvis_tts_perf_client.cc.

jarvis_tts_perf_client --text_file=/work/test_files/tts/ljs_audio_text_test_filelist_small.txt

Python Client Examples¶

Two command line clients implemented in Python are provided as an example of how to interact with the TTS service using the Python API. talk.py is an implementation of a client for performing offline inference, and talk_stream.py an implementation using online inference (with audio being streamed back to the client in chunks).