Question About bit_config.py

Zhen-Dong / HAWQ

Quantization library for PyTorch. Support low-precision and mixed-precision quantization, with hardware implementation through TVM.

MIT License

406 stars 83 forks source link

Question About bit_config.py #39

Open rainyBJ opened 1 year ago

rainyBJ commented 1 year ago

Great work! I'm wondering why the re-quant op before stage1 is always configured to 16 bits as 'quant_act_int32': 16. From my perspective, configure it to the same as 'stage1.unit1.quant_act' seems to make no difference.