reviewer feedback

Author: sayakpaul

src/diffusers/models/modeling_utils.py

@@ -1550,7 +1550,6 @@ def _load_pretrained_model(
         # tensors using their expected shape and not performing any initialization of the memory (empty data).
         # When the actual device allocations happen, the allocator already has a pool of unused device memory
         # that it can re-use for faster loading of the model.
-        # TODO: add support for warmup with hf_quantizer
         if device_map is not None:
             expanded_device_map = _expand_device_map(device_map, expected_keys)
             _caching_allocator_warmup(model, expanded_device_map, dtype, hf_quantizer)