in optimum/quanto/tensor/activations/qbytes_ops.py [0:0]
def where(op, condition, input, other):
if isinstance(condition, QTensor) or isinstance(other, QTensor):
raise NotImplementedError
float_data = op(condition, input.dequantize(), other)
if input.axis is None:
# We requantize with the input scale
return quantize_activation(float_data, qtype=input.qtype, scale=input._scale)
return float_data