Incompatibility between transformers 4.45.0 and torch 1.9.1

[realjoshqsun]

System Info

Hello, I was using depth estimation model,
pipe = pipeline(task="depth-estimation", model="depth-anything/Depth-Anything-V2-Metric-Indoor-Small-hf")
But I got this error:

Traceback (most recent call last):
  File "test.py", line 3, in <module>
    pipe = pipeline(task="depth-estimation", model="depth-anything/Depth-Anything-V2-Metric-Indoor-Small-hf")
  File "/home/q84sun/miniconda3/envs/vlnce_py3.8/lib/python3.8/site-packages/transformers/pipelines/__init__.py", line 896, in pipeline
    framework, model = infer_framework_load_model(
  File "/home/q84sun/miniconda3/envs/vlnce_py3.8/lib/python3.8/site-packages/transformers/pipelines/base.py", line 288, in infer_framework_load_model
    model = model_class.from_pretrained(model, **kwargs)
  File "/home/q84sun/miniconda3/envs/vlnce_py3.8/lib/python3.8/site-packages/transformers/models/auto/auto_factory.py", line 564, in from_pretrained
    return model_class.from_pretrained(
  File "/home/q84sun/miniconda3/envs/vlnce_py3.8/lib/python3.8/site-packages/transformers/modeling_utils.py", line 3808, in from_pretrained
    state_dict = load_state_dict(resolved_archive_file)
  File "/home/q84sun/miniconda3/envs/vlnce_py3.8/lib/python3.8/site-packages/transformers/modeling_utils.py", line 556, in load_state_dict
    return safe_load_file(checkpoint_file)
  File "/home/q84sun/miniconda3/envs/vlnce_py3.8/lib/python3.8/site-packages/safetensors/torch.py", line 315, in load_file
    result[k] = f.get_tensor(k)
AttributeError: module 'torch' has no attribute 'frombuffer'

It seemed like a compatible issue between transformers and torch. What is the right torch version to match transformers 4.45.0?

My environment:
ubuntu: 22.04
Python: 3.8.20
torch: 1.9.1+cu111
transformers: 4.45.0
nvcc: cuda_11.7

Who can help?

@amyeroberts @qubvel @Rocketknight1

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

I was using the official example scripts from https://huggingface.co/depth-anything/Depth-Anything-V2-Metric-Indoor-Small-hf

from transformers import pipeline

pipe = pipeline(task="depth-estimation", model="depth-anything/Depth-Anything-V2-Metric-Indoor-Small-hf")

When running this test script with the my own environment, it raised up an error.

Expected behavior

Should be working.