in aidial_adapter_vertexai/embedding/multi_modal.py [0:0]
def count_input_tokens(self) -> int:
# The model doesn't report the number of input tokens.
# However, one could count it oneself:
# https://cloud.google.com/vertex-ai/generative-ai/pricing#embedding-models
# As of 29 Jul 2024, one image costs as much as 500 text input characters
ret = len(self.contextual_text or "")
if self.image:
ret += 500
return ret