File size: 4,032 Bytes
3215d8d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 |
import tritonclient.http as http_client
from tritonclient.utils import *
DEFAULT_SAMPLING_RATE = 22050
ENABLE_SSL = False
ENDPOINT_URL = 'localhost:8000'
HTTP_HEADERS = {"Authorization": "Bearer __PASTE_KEY_HERE__"}
# Connect to the server
if ENABLE_SSL:
import gevent.ssl
triton_http_client = http_client.InferenceServerClient(
url=ENDPOINT_URL, verbose=False,
ssl=True, ssl_context_factory=gevent.ssl._create_default_https_context,
)
else:
triton_http_client = http_client.InferenceServerClient(
url=ENDPOINT_URL, verbose=False,
)
print("Is server ready - {}".format(
triton_http_client.is_server_ready(headers=HTTP_HEADERS)
))
import io
from scipy.io.wavfile import write as scipy_wav_write
import numpy as np
def get_string_tensor(string_value, tensor_name):
string_obj = np.array([string_value], dtype="object")
input_obj = http_client.InferInput(tensor_name, string_obj.shape, np_to_triton_dtype(string_obj.dtype))
input_obj.set_data_from_numpy(string_obj)
return input_obj
# Uncomment based on the language you want to test
## North-Indian languages
# inputs = [get_string_tensor("নমস্তে", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("as", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("নমস্তে", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("bn", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("નમસ્તે", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("gu", "INPUT_LANGUAGE_ID")]
inputs = [get_string_tensor("सलाम", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("hi", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("नमस्ते", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("mr", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("ନମସ୍ତେ", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("or", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("ਨਮਸਤੇ", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("pa", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("सलाम", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("raj", "INPUT_LANGUAGE_ID")]
## South-Indian languages
# inputs = [get_string_tensor("ನಮಸ್ಕಾರಂ", "INPUT_TEXT"), get_string_tensor("male", "INPUT_SPEAKER_ID"), get_string_tensor("kn", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("നമസ്കാരം", "INPUT_TEXT"), get_string_tensor("male", "INPUT_SPEAKER_ID"), get_string_tensor("ml", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("வணக்கம்", "INPUT_TEXT"), get_string_tensor("male", "INPUT_SPEAKER_ID"), get_string_tensor("ta", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("నమస్కారం", "INPUT_TEXT"), get_string_tensor("male", "INPUT_SPEAKER_ID"), get_string_tensor("te", "INPUT_LANGUAGE_ID")]
## Misc languages
# inputs = [get_string_tensor("Greetings", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("en", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("नमस्ते", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("brx", "INPUT_LANGUAGE_ID")]
# inputs = [get_string_tensor("নমস্তে", "INPUT_TEXT"), get_string_tensor("female", "INPUT_SPEAKER_ID"), get_string_tensor("mni", "INPUT_LANGUAGE_ID")]
output0 = http_client.InferRequestedOutput("OUTPUT_GENERATED_AUDIO")
response = triton_http_client.infer(
"tts",
model_version='1',
inputs=inputs,
outputs=[output0],
headers=HTTP_HEADERS,
)#.get_response()
raw_audio = response.as_numpy("OUTPUT_GENERATED_AUDIO")[0]
byte_io = io.BytesIO()
scipy_wav_write(byte_io, DEFAULT_SAMPLING_RATE, raw_audio)
with open("audio.wav", "wb") as f:
f.write(byte_io.read())
|