Update checkpoint.py for quantization_bit setting

analogdevicesinc / ai8x-synthesis

Quantization and Synthesis (Device Specific Code Generation) for ADI's MAX78000 and MAX78002 Edge AI Devices

Apache License 2.0

55 stars 47 forks source link

Update checkpoint.py for quantization_bit setting #285

Closed seldauyanik-maxim closed 1 year ago

seldauyanik-maxim commented 1 year ago

Checkpoint.py was determining quantization_bits per layer automatically using layer weights' interval resulting in a faulty value for some samples. This sections are updated to use weight_bits instead.