Nice work accelerating convolutional models! It might be better to see (or cite papers about) the trade-off how model performance (accuracy, etc) changes w.r.t.

Nice work accelerating convolutional models! It might be better to see (or cite papers about) the trade-off how model performance (accuracy, etc) changes w.r.t. how it is quantized.