From 5dce71010174ad6599653da4e8ba37fd5f9fa572 Mon Sep 17 00:00:00 2001 From: Ross Wightman Date: Mon, 27 May 2024 14:56:03 -0700 Subject: [PATCH] Add vit_little in12k + in12k-ft-in1k weights --- timm/models/vision_transformer.py | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) diff --git a/timm/models/vision_transformer.py b/timm/models/vision_transformer.py index 9c95ae4b..2dd4754a 100644 --- a/timm/models/vision_transformer.py +++ b/timm/models/vision_transformer.py @@ -1872,6 +1872,13 @@ default_cfgs = { 'vit_pwee_patch16_reg1_gap_256.sbb_in1k': _cfg( hf_hub_id='timm/', input_size=(3, 256, 256), crop_pct=0.95), + 'vit_little_patch16_reg1_gap_256.sbb_in12k_ft_in1k': _cfg( + hf_hub_id='timm/', + input_size=(3, 256, 256), crop_pct=0.95), + 'vit_little_patch16_reg1_gap_256.sbb_in12k': _cfg( + hf_hub_id='timm/', + num_classes=11821, + input_size=(3, 256, 256), crop_pct=0.95), 'vit_little_patch16_reg4_gap_256.sbb_in1k': _cfg( hf_hub_id='timm/', input_size=(3, 256, 256), crop_pct=0.95), @@ -2975,6 +2982,17 @@ def vit_pwee_patch16_reg1_gap_256(pretrained: bool = False, **kwargs) -> VisionT return model +@register_model +def vit_little_patch16_reg1_gap_256(pretrained: bool = False, **kwargs) -> VisionTransformer: + model_args = dict( + patch_size=16, embed_dim=320, depth=14, num_heads=5, init_values=1e-5, mlp_ratio=5.6, + class_token=False, no_embed_class=True, reg_tokens=1, global_pool='avg', + ) + model = _create_vision_transformer( + 'vit_little_patch16_reg1_gap_256', pretrained=pretrained, **dict(model_args, **kwargs)) + return model + + @register_model def vit_little_patch16_reg4_gap_256(pretrained: bool = False, **kwargs) -> VisionTransformer: model_args = dict(