search

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

YAML file #261

Closed salihtangel closed 1 year ago

salihtangel commented 1 year ago

In synthesies i get this error "Layer 8: 3 input channels (before flattening) using 1 pass, and 1 operand (3 processors per pass), but the enabled processor map 0x00000000ffffffff has 32 bits instead of the expected number of 3." I could not solved it. Can you share a yaml file for ai85kws20netv2batchnorm model?

aniktash commented 1 year ago

Make sure to use the same model for training and synthesis. The v3 model is used in demo application, which is much smaller than v2. If you want to use v2 model with batchnorm: 1- In training and evaluate v2 script, change the model and add bias: --model ai85kws20netv2batchnorm --use-bias 2- Copy the kws20-v2-hwc.yaml as kws20-v2-batchnorm-hwc.yaml 3- In kws20-v2-batchnorm-hwc.yaml file, change arch name to match the batchnorm model: arch: ai85kws20netv2batchnorm 4- in synthesis script use this new yaml: --config-file networks/kws20-v2-batchnorm-hwc.yaml