search

Xilinx / Vitis-AI

Vitis AI is Xilinx’s development stack for AI inference on Xilinx hardware platforms, including both edge devices and Alveo cards.
https://www.xilinx.com/ai
Apache License 2.0
1.43k stars 621 forks source link

vitis fast finetune #563

Open EliyaBen opened 2 years ago

EliyaBen commented 2 years ago

Post-training quantization (PTQ) - without finetune and Quantization aware training (QAT) works fine but get error in Post-training quantization (PTQ) - fast finetune: activation = layer.layer.activiation.activiation AttributeError: 'Concatenate' object has no attribute 'activation'

vitis-ai 1.4 network: densenetx

xieyi4650 commented 2 years ago

which framework are you using?

xieyi4650 commented 2 years ago

Hi @EliyaBen could you please provide a simple model to reproduce the error.