Skip to content

No backbone for CLS #33

@BodongDu

Description

@BodongDu

python main.py experiment=task/$task data=Cls/$dataset data="{batch_size:$batch_size,num_workers:$num_workers}"
model=$model model.model_cfg.backbone.pretrained=$pretrained_path
train="{epochs:$epochs, accumulation_steps:1}" L="{devices:$devices}" tag=USFM
I think you should change to:
python main.py experiment=task/$task data=Cls/$dataset data="{batch_size:$batch_size,num_workers:$num_workers}"
model=$model model.model_cfg.pretrained=$pretrained_path
train="{epochs:$epochs, accumulation_steps:1}" L="{devices:$devices}" tag=USFM
I have another question: Whether the pretrained weight could be used for CLS.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions