[AMD ROCM] Encountered exception while importing vector_quantize_pytorch: cannot import name 'group' from 'torch.distributed'

[cwayne1989]

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:
Follow the Windows Portable Build using the Requirements.ROCM.TXT file install instructions.

Expected behavior
The model to load.

Screenshots

Windows 11 x 64
Amd Radeon 9060XT

[coderabbitai]

⚠️ Possible Duplicate Issue(s)

• ROCM7.2 vector_quantize_pytorch: cannot import name 'group' from 'torch.distributed' #561

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[LeeAeron]

Describe the bug A clear and concise description of what the bug is.

To Reproduce Steps to reproduce the behavior: Follow the Windows Portable Build using the Requirements.ROCM.TXT file install instructions.

Expected behavior The model to load.

Screenshots

Windows 11 x 64 Amd Radeon 9060XT

check models folder / turbo model for modeling_acestep_v15_turbo.py file. It seems your model broken or not downloaded properly. delete turbo model folder and re-start Ace - it must download model again.
Try to change source from Hugginface to Modelscope, if problems with downloading from HugginFace: in .bat file change line to
set DOWNLOAD_SOURCE=--download-source modelscope

If model is OK, then it seems not compatible Transformers version in requirements.txt. check installed Transformers and if not version as pointed in requirements reinstall proper version as is in requirements.txt