[Feature request] Image-to-Image for super-resolution

[xenova]

Feature request

Running

optimum-cli export onnx -m caidas/swin2SR-realworld-sr-x4-64-bsrgan-psnr out

Results in

Framework not specified. Using pt to export to ONNX.
Traceback (most recent call last):
  File "/usr/local/bin/optimum-cli", line 8, in <module>
    sys.exit(main())
  File "/usr/local/lib/python3.10/dist-packages/optimum/commands/optimum_cli.py", line 163, in main
    service.run()
  File "/usr/local/lib/python3.10/dist-packages/optimum/commands/export/onnx.py", line 225, in run
    main_export(
  File "/usr/local/lib/python3.10/dist-packages/optimum/exporters/onnx/__main__.py", line 306, in main_export
    model = TasksManager.get_model_from_task(
  File "/usr/local/lib/python3.10/dist-packages/optimum/exporters/tasks.py", line 1481, in get_model_from_task
    if TasksManager._TASKS_TO_LIBRARY[task.replace("-with-past", "")] == "transformers":
KeyError: 'image-to-image'

meaning this task is not yet supported. I also tried selecting the masked-im task, however that results in

Framework not specified. Using pt to export to ONNX.
Traceback (most recent call last):
  File "/usr/local/bin/optimum-cli", line 8, in <module>
    sys.exit(main())
  File "/usr/local/lib/python3.10/dist-packages/optimum/commands/optimum_cli.py", line 163, in main
    service.run()
  File "/usr/local/lib/python3.10/dist-packages/optimum/commands/export/onnx.py", line 225, in run
    main_export(
  File "/usr/local/lib/python3.10/dist-packages/optimum/exporters/onnx/__main__.py", line 306, in main_export
    model = TasksManager.get_model_from_task(
  File "/usr/local/lib/python3.10/dist-packages/optimum/exporters/tasks.py", line 1513, in get_model_from_task
    model = model_class.from_pretrained(model_name_or_path, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/transformers/models/auto/auto_factory.py", line 558, in from_pretrained
    raise ValueError(
ValueError: Unrecognized configuration class <class 'transformers.models.swin2sr.configuration_swin2sr.Swin2SRConfig'> for this kind of AutoModel: AutoModelForMaskedImageModeling.
Model type should be one of DeiTConfig, FocalNetConfig, SwinConfig, Swinv2Config, ViTConfig.

See here for a list of available swin2sr models: https://huggingface.co/models?other=swin2sr

Motivation

Requested here by @josephrocca. With WebGPU support right around the corner, I'd like to have a variety of vision models ready to create demos for (on top of, for example, SAM which is now supported).

Your contribution

I will add support to transformers.js when ready