Spaces:

eduardzamfir

/

SeeMoreDetails

like 184

Running

This PR simplifies the loading of the model from the hub, along with working download metrics.

Basically, when the model is defined like so:

from huggingface_hub import hf_hub_download

# define model
model = seemore.SeemoRe(scale=cfg.model.scale, in_chans=cfg.model.in_chans,
                        num_experts=cfg.model.num_experts, num_layers=cfg.model.num_layers, embedding_dim=cfg.model.embedding_dim, 
                        img_range=cfg.model.img_range, use_shuffle=cfg.model.use_shuffle, global_kernel_size=cfg.model.global_kernel_size, 
                        recursive=cfg.model.recursive, lr_space=cfg.model.lr_space, topk=cfg.model.topk)
# load weights
filepath = hf_hub_download(repo_id="eduardzamfir/SeemoRe-T", filename="SeemoRe_T_X4.pth", local_dir="./")
model.load_state_dict(filepath, map_location="cpu")

You can do things like:

# push to the hub
model.push_to_hub("eduardzamfir/SeemoRe-T")
# save to a local folder
model.save_pretrained(".")
# load from the hub
model.from_pretrained("eduardzamfir/SeemoRe-T")

This would ensure that download metrics work, along with better tags etc.

See the PyTorchModelHubMixin docs for more info: https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin