robotsp

did you fix it? @nikhiljaiswal

@AIikai I have a similar question. Do you have any ideas?

@SefaZeng Would you please provide your code? I have the same question.

@YingChenlu I have a similar question, do you know how to freeze parameters of the model in fairseq when training. I used both zero_grad and requires_grad = false are not...

> @gmryu Thanks. Yes, I think so. But I have tried and **it raises an error when I reloaded the model**: args = state["args"] if arg_overrides is not None: for...

> @robotsp I guess that will happen since a model state dict is not a checkpoint for fairseq. I found you can use torch.save instead. The flow is you load...

> I guess at this point, you do not load as a model but load the checkpoint itself. @gmryu Thanks for your reply, I am going to finetune the new...

> Thanks @gmryu , it works. Btw, do you know any methods of pruning (L1 pruning, etc..) fairseq models?

@gmryu I have a try using Microsoft NNI to prune fairseq model, and here is the code, ` import torch from torch import nn from torchvision import models from torch...

@gmryu I tried to assign tensors of random integers as input and it raises an error at `ModelSpeedup(model, dummy_input=torch.rand([64,50,50,64,50]), masks_file=masks).speedup_model()` error msg: ``` NotImplementedError Traceback (most recent call last) in...