A rather tiny PR, "fixes" or more likely errors properly on an issue introduced in #3628. The issue is that DDP doesn't work if model has DTensors which it has if it was loaded with device_map="auto" (it applies TP). Now we properly raise an Error.

Snippet to test:

from transformers import AutoModelForCausalLM

from accelerate import Accelerator


model_id = "NousResearch/Hermes-3-Llama-3.1-8B"
accelerator = Accelerator()

model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

model = accelerator.prepare(model)

Before:
[rank0]: AssertionError: found no DeviceMesh from dtensor args for c10d.broadcast_.default!

After:
[rank0]: ValueError: Your model contains DTensorparameters, which is incompatible with DDP. Maybe you loaded your model withdevice_map='auto'?

cc @SunMarc