Fix --local-rank argument compatibility with torchrun

[Mr-Neutr0n]

Summary

• Accept both --local-rank (hyphen) and --local_rank (underscore) for the distributed training argument
• Fixes compatibility with PyTorch's torchrun which passes --local-rank instead of --local_rank

Problem

When using torchrun for distributed training:

torchrun --nproc_per_node=6 train.py ...

The script fails with:

HQ-SAM: error: unrecognized arguments: --local-rank=5

This happens because torchrun passes --local-rank (with hyphen), but the script only accepted --local_rank (with underscore).

Solution

Changed the argument definition to accept both forms using argparse's multiple option names feature.

Test plan

• Run distributed training with torchrun --nproc_per_node=N train.py
• Verify backward compatibility with torch.distributed.launch (deprecated)

Fixes #153

[Mr-Neutr0n]

bump — this fixes the --local-rank argument to work with torchrun's --local_rank convention. otherwise it crashes on launch. lmk if anything needs changing

[Mr-Neutr0n]

Friendly ping -- this one-line fix in train/train.py resolves the --local-rank vs --local_rank mismatch that causes torchrun to crash on launch. Would appreciate a review when you get a chance.

[Mr-Neutr0n]

Closing this one to tidy up my open pull requests.

It's been open around five months with no review activity, and I'd rather withdraw it than leave it sitting in your queue. Nothing needed from you, and no hard feelings at all.

If the fix is still wanted, this can be reopened, or I'm happy to redo it properly against current main. Apologies for the noise.