Open nikky4D opened 3 years ago
I can follow the format in hubconf.py. Is the backbone for the LVIS model also a Resnet101? I'm making something like this, however it is throwing errors:
`
def mdetr_resnet101_lvis(pretrained=False, return_postprocessor=False):
model = _make_detr("resnet101")
if pretrained:
checkpoint = torch.hub.load_state_dict_from_url(
url="https://zenodo.org/record/4721981/files/lvis10_checkpoint.pth",
map_location="cpu",
check_hash=True,
)
model.load_state_dict(checkpoint["model"])
if return_postprocessor:
return model, PostProcess()
return model
`
Hi,
Is there a sample on how to load the lvis model in colab, similar to what is currently done with torch.hub?