[С++] Updating С++ examples

[snakers4]

Dear @OJRYK and @NathanJHLee,

As per popular request, could you maybe help with updating your cpp examples for the actual version?

In v5 we slightly changed the inputs.

[snakers4]

#2 (comment) these changes were introduced, probably the easiest way to see them is to compare commits related to V5 to the previous state

v6 did bot really change the inputs, just some minor improvements that can be skipped

[snakers4]

window_size_samples is deprecated - now the VAD only works with fixed size window;

VAD now works with 8 kHz and 16 kHz sample rates, only with fixed 256 and 512 sample windows respectively;

Slightly changed internal logic, now some context (part of previous chunk) is passed along with the current chunk;

[humblemh]

"Looks like the sample code might have a bug. In the VadIterator constructor, it seems the following line is missing:
speech_pad_samples = sr_per_ms * speech_pad_ms;
Could you check if this needs to be added?"

silero-vad-onnx.cpp

[NathanJHLee]

Hello @snakers4
Sorry for the late response. I understand that only the fixed window_size_samples values change depending on the sample rate, and that Silero-VAD supports only 16 kHz and 8 kHz.
Did I understand it correctly?

I think it’s only a tiny change, so I’ll push a new version in several days
The new version will also support the ONNX model.

[snakers4]

Sorry for the late response. I understand that only the fixed window_size_samples values change depending on the sample rate, and that Silero-VAD supports only 16 kHz and 8 kHz.
Did I understand it correctly?

Yes, this is correct.

Please also note that in new versions some context is also concatenated to the left (ONNX example):

silero-vad/src/silero_vad/utils_vad.py

Lines 57 to 92 in be95df9

	def __call__(self, x, sr: int):
	x, sr = self._validate_input(x, sr)
	num_samples = 512 if sr == 16000 else 256
	if x.shape[-1] != num_samples:
	raise ValueError(f"Provided number of samples is {x.shape[-1]} (Supported values: 256 for 8000 sample rate, 512 for 16000)")
	batch_size = x.shape[0]
	context_size = 64 if sr == 16000 else 32
	if not self._last_batch_size:
	self.reset_states(batch_size)
	if (self._last_sr) and (self._last_sr != sr):
	self.reset_states(batch_size)
	if (self._last_batch_size) and (self._last_batch_size != batch_size):
	self.reset_states(batch_size)
	if not len(self._context):
	self._context = torch.zeros(batch_size, context_size)
	x = torch.cat([self._context, x], dim=1)
	if sr in [8000, 16000]:
	ort_inputs = {'input': x.numpy(), 'state': self._state.numpy(), 'sr': np.array(sr, dtype='int64')}
	ort_outs = self.session.run(None, ort_inputs)
	out, state = ort_outs
	self._state = torch.from_numpy(state)
	else:
	raise ValueError()
	self._context = x[..., -context_size:]
	self._last_sr = sr
	self._last_batch_size = batch_size
	out = torch.from_numpy(out)
	return out

The PyTorch model hanles this inside of the model.

[snakers4]

#721

Updates for libtorch and onnx-runtime cpp examples