accelerator.prepare

[wfan1203]

I want to use data files from different folders for each rank. What should I do? Do I need to add some parameters when using accelerator.prepare?

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[wfan1203]

感谢您的来信，我将及时回复您！                                                   ——武凡

[Sakura1a2a3a]

I also met an issue with accelerator.prepare method.
I have a pretrained model and I want it only do the inference without any parameter changes. I put that model = accelerator.prepare(model), such as this and I find the outputs of the model before and after accelerator.prepare() with the same input are different. Could you please explain what is going on?

The mix_precision = "no" and split_batches = True.

[wfan1203]

感谢您的来信，我将及时回复您！                                                   ——武凡

[muellerzr]

During inference you should not prepare the model, as our inference examples show: https://github.com/huggingface/accelerate/blob/main/examples/inference/distributed/phi2.py

You can/should use accelerate to prepare the dataloader only

[Sakura1a2a3a]

Oh great! So there is something different before and after the prepare() function, just what to make sure? What about there are 2 models, one for training another for inference; In this case should I only use the accelerator for training model and dataloader but leave the inference model out of it?

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[wfan1203]

感谢您的来信，我将及时回复您！                                                   ——武凡

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[wfan1203]

感谢您的来信，我将及时回复您！                                                   ——武凡

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[wfan1203]

感谢您的来信，我将及时回复您！                                                   ——武凡

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[wfan1203]

感谢您的来信，我将及时回复您！                                                   ——武凡

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.