Can the model be trained simultaneously using quantization and sparsity?

[Vieeo]

Any examples ？

[ChenhanYu]

The short answer is yes but we didn't maintain an example. You should be able to add these lines to apply simulated sparsity after quantized the model using modelopt.torch.quantize. The model can still be trained.