Skip to content

Update unquantized pertained model

Latest
Compare
Choose a tag to compare
@hamling-ling hamling-ling released this 06 May 04:55

Unquantized version of pertained model
works with rev f91cdcc

Including checkpoints of training without --enable_quantization option.

pretrained_model-1.3.zip includes result of training WITHOUT --enable_quantization option
frozen_model-1.3.zip includes inference graph, frozen graph and model exported to fp16 tflite file.