Add support for model training from checkpoint

[Littleor]

This pull request adds support for model training continuation from a checkpoint, making it possible to resume the training process at a later point in time.

NOTE:
Please note that this pull request adds support for model training from a checkpoint because the previous pull request (#23) did not work properly in my environment.

To utilize this feature, simply include the --ckpt parameter, such as:

torchrun --nnodes=1 --nproc_per_node=N train.py --model DiT-XL/2 --data-path /path/to/imagenet/train --ckpt /path/to/model.pt

[facebook-github-bot]

Hi @Littleor!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[chenllliang]

Hi @Littleor I met a cuda-out-of-memory error after loading from a checkpoint before training. Do you have any idea why this happened? Training from scratch is ok.

[forever208]

Hi @Littleor I met a cuda-out-of-memory error after loading from a checkpoint before training. Do you have any idea why this happened? Training from scratch is ok.

same here, @chenllliang have you solved it?

[Chen-yu-Zheng]

Hi @Littleor I met a cuda-out-of-memory error after loading from a checkpoint before training. Do you have any idea why this happened? Training from scratch is ok.

I just change one line code: checkpoint = torch.load(args.ckpt, map_location=lambda storage, loc: storage) and everything looks good.

[Littleor]

Hi @Littleor I met a cuda-out-of-memory error after loading from a checkpoint before training. Do you have any idea why this happened? Training from scratch is ok.

I just change one line code: checkpoint = torch.load(args.ckpt, map_location=lambda storage, loc: storage) and everything looks good.

Yes, this is the source of the problem. Loading chekpoint needs to be done on the appropriate device to avoid OOM.