PyTorch EfficientNet Lite
quic-bharathr
released this
31 Dec 07:47
·
28 commits
to develop
since this release
This is the PyTorch EfficientNet Lite optimized checkpoint
- Batch Norm folding and Adaround has been applied on the original model
- Adaround has been optimized with 8-bit width and "tf_enhanced" quant scheme
- Quantization evaluated with "tf_enhanced" quant scheme in 8 bit width weight & activation quantization