How to use the retrival large model for image-text prediction (model_large_retrieval_coco) ?

[caydenwei]

In the Image-Text Matching part in Colab, I try to use the large retrieval coco model provided in README.
But it reports an Assertion error as follow, what can I do to do the prediction using model_url: https://storage.googleapis.com/sfr-vision-language-research/BLIP/models/model_large_retrieval_coco.pth

Traceback (most recent call last):
  File "/app-dev/rootfs/app/main.py", line 144, in <module>
    run_tasks(args.config, args.task)
  File "/app-dev/rootfs/app/main.py", line 103, in run_tasks
    clip = CLIP(pipeline=pipeline_name, impl=impl, model_name=model, **kwargs)
  File "/app-dev/rootfs/app/ops_clip.py", line 20, in __init__
    _, self.preprocess = self.clip.load_model()
  File "/app-dev/rootfs/app/framework/impl/clip/blip_predictor.py", line 30, in load_model
    model = blip_itm(med_config=os.path.join(blip_path, 'configs/med_config.json'), pretrained=self.model_url,
  File "/app-dev/rootfs/app/framework/impl/clip/BLIP/models/blip_itm.py", line 74, in blip_itm
    assert(len(msg.missing_keys)==0)
AssertionError