Add example of amp training & qunatization-aware training

[nieyan]

[qubvel]

Hey, thanks for opening the issue, can you provide more details in the description?

[nieyan]

I used mobileone backbone and it was hard to quantization to int8. What's more, the training speed seem too slow if haven't amp .

[qubvel]

Thanks for the details. Did you succeed in training and quantization to uint8? If so, would you like to contribute a PR with an example? I don't have such experience, so I would greatly appreciate the contribution.

[nieyan]

Thanks for the details. Did you succeed in training and quantization to uint8? If so, would you like to contribute a PR with an example? I don't have such experience, so I would greatly appreciate the contribution.

unfortunately, I haven't find the soultion so I just using mix quantizaiton and that was slow in embedded device.

[qubvel]

Ok, let's see if anyone else has a solution