don't load state_dict twice when using low_cpu_mem_usage in from_pretrained

[patil-suraj]

What does this PR do?

In from_pretrained the state_dict is loaded twice when low_cpu_mem_usage is True, which is not required since the state_dict is already loaded before as we can here.

transformers/src/transformers/modeling_utils.py

Lines 1795 to 1797 in 21decb7

	if not is_sharded:
	# Time to load the checkpoint
	state_dict = load_state_dict(resolved_archive_file)

So, it's fine to remove it there because when is_sharded=False, the state_dict is loaded at line 1797

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[sgugger]

To complete the description, it's fine to remove it there because when is_sharded=False, the state_dict is loaded at line 1797.