SR_Mobile_Quantization

SR_Mobile_Quantization copied to clipboard

RuntimeError: Quantization not yet supported for op: 'DEPTH_TO_SPACE'. I would appreciate it if you give me some solutions about this problem.

biuping

Nice job! I am confused about the difference between the tf.yaml and tfnightly.yaml? Does the key difference is the version of tensorflow? If this is true, can I only use...

coolKeen

it seems the inference is very slow on my linux server?

5

Hi, Dear NJU-Jet my linux server: several 2.6GHz CPU + several V100, and I run the **generate_tflite.py** to got a quantized model. and then in function **evaluate**, I add below...

xiaoxiongli

I want to train the model and then I need a int8 onnx model, what should I do step by step?

6

Dear @NJU-Jet : I feel confused that if I use the command "**python train.py --opt options/train/base7.yaml --name base7_D4C28_bs16ps64_lr1e-3 --scale 3 --bs 16 --ps 64 --lr 1e-3 --gpu_ids 0**" to train...

xiaoxiongli

running time on AI Benchmark App

3

Hello: I tested 'base7_D4C28_bs16ps64_lr1e-3_qat_time.tflite' running time via AI Benchmark App. My device is Snapdragon 888 and the device's AI score is 54.4. It takes about 200ms NNAPI. In the paper,...

caijie0620

Would be great if you could share the code or explain how to investigate for other devices

nisargshah1999

NJU-Jet

How to deal with QuantizeLinear and DequantizeLinear node when I do qutization using openvino/tnn/mnn?

1

I train a x2 model, and after that I finetune it using QAT training using below command: >"python train.py --opt options/train/base7_qat.yaml --name base7_D4C28_bs16ps64_lr12-3_qat_x2 --scale 2 --bs 16 --ps 64 --lr...

xiaoxiongli

During QAT training for scale = 2, do I need change below "3" to "2" ?

1

I have already trained a model that scale = 2, and after that I wanna to do QAT training for it(scale = 2) using command "**python train.py --opt options/train/base7_qat.yaml --name...

xiaoxiongli

tensorflow version

2

Thanks for sharing your great work. I have a question about TF version. Can I derectly only use TF 2.5.0 instead of TF2.4 and tf-nightly=2.5?

caijie0620