The provided checkpoint is not working

[szlll]

when I loaded the model with the provided checkpoint, I found the following error:
"module 'models' has no attribute 'ICLIP_VITB16'. Did you mean: 'ACLIP_VITB16'?"

Here is my code to load the model:

ckpt_path = "/path/to/provided/checkpoint_best.pt"
ckpt = torch.load(ckpt_path, map_location='cpu')
state_dict = OrderedDict()
for k, v in ckpt['state_dict'].items():
state_dict[k.replace('module.', '')] = v

old_args = ckpt['args']
print("=> creating model: {}".format(old_args.model))
detailclip = getattr(models, old_args.model)()
detailclip.cuda()
detailclip.load_state_dict(state_dict, strict=True)
print("=> loaded resume checkpoint '{}' (epoch {})".format(args.resume, ckpt['epoch']))

[rleygonie]

I have the same error. Did you manage to get around the problem, or is the checkpoint really the wrong one?

[KishoreP1]

Sorry for the late response. We renamed the model to "DetailCLIP_VITB16" from "ICLIP_VITB16" before publishing the repo.

Try renaming as shown below:

# ... (your code to load ckpt and state_dict)
old_args = ckpt['args']

# ADD THIS LINE:
old_args.model = 'DetailCLIP_VITB16'

# The rest of your code can stay the same:
print("=> creating model: {}".format(old_args.model))
detailclip = getattr(models, old_args.model)()
# ... (the rest of your script)