MACE supports post-training quantization if you want to take a chance to quantize model directly without fine tuning.
This method requires developer to calculate tensor range of each activation layer statistically using sample inputs.
MACE provides tools to do statistics with following steps:
MACE provides tools to do statistics with following steps(using `inception-v3` from `MACE Model Zoo <https://github.com/XiaoMi/mace-models>`__ as an example):
1. Convert original model to run on CPU host without obfuscation (by setting `target_abis` to `host`, `runtime` to `cpu`,
and `obfuscate` to `0`, appending `:0` to `output_tensors` if missing in yaml config).
and `obfuscate` to `0`, appending `:0` to `output_tensors` if missing in yaml config).