Quantizing the models using tflite

General Discussion
, , ,
#1

Hi Everyone, I am trying to quantize my model to 4-bit using post-training quantization, I see tflite supports int 8 bits and float 16 bit. How can do post-training quantization to convert my float 32-bit model to int 4-bit model . Is it possible ? Kindly provide your views

Regards
Saras

#3

Hi @saras26, Currently there is no support for int 4 quantization. Only int8, float16 are minimum quantization supported. Thank You.

1 Like
#5

Thank you so much for the clarification.