Thai Nguyen

icon indicating a website link thaink.github.io icon indicating an email [email protected]

icon company @Google icon location Seoul, Korea icon location Software Engineer.

Results 22 comments of Thai Nguyen

Full Int8 QAT not working

Hi Matty, Passing `converter.representative_dataset = representative_dataset` is only required for post-training quantization. If you want to use QAT, follow the guide at https://www.tensorflow.org/model_optimization/guide/quantization/training_example( use `quantize_model` before training and train it...

Full Int8 QAT not working

The inference_input_type and inference_output_type is to use int8 input and output actually.

Full Int8 QAT not working

Can you share or describe what your output model looks like?

Full Int8 QAT not working

@Xhark Could you check if the Matt is QAT-ing the right way?

Unsupported operations when applying tfmot

@Xhark Could you take a look.

Quantization aware training for a Transfer Learning MobileNet model

@Xhark could you take a look at this issue?

Pruning only works for small batch sizes

@rino20 Could you take a look?

How to extract original model's weight from trained QAT model?

Isn't the original model weights in checkpoint files or variables/ directory?

How to extract original model's weight from trained QAT model?

I don't think we have an official way for that. This use with TensorRT isn't in our set of use-cases for now.

bilinear upsampling layer cannot be 8-bit quantized correctly

Jaehong is better in QAT and Keras-related things. @Xhark Could you take a look?