mirrors/pytorch-image-models
1
0
Fork 0
mirror of https://github.com/huggingface/pytorch-image-models.git synced 2025-06-03 15:01:08 +08:00
Code Issues Projects Releases Wiki Activity

Remove duplicate so400m/16 @ 256 model def

Browse Source
This commit is contained in:
Ross Wightman 2025-02-20 16:14:58 -08:00 committed by Ross Wightman
parent 9758e0b8b0
commit f63a11cf81
1 changed files with 3 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
timm/models/vision_transformer.py
View File

@ -3499,18 +3499,6 @@ def vit_so400m_patch14_siglip_gap_224(pretrained: bool = False, **kwargs) -> Vis
return model return model
@register_model
def vit_so400m_patch16_siglip_gap_256(pretrained: bool = False, **kwargs) -> VisionTransformer:
""" A SigLIP variant of ViT with global average pooling (GAP) instead of attention pooling (MAP)."""
model_args = dict(
patch_size=16, embed_dim=1152, depth=27, num_heads=16, mlp_ratio=3.7362,
class_token=False, global_pool='avg', fc_norm=False,
)
model = _create_vision_transformer(
'vit_so400m_patch16_siglip_gap_256', pretrained=pretrained, **dict(model_args, **kwargs))
return model
@register_model @register_model
def vit_so400m_patch14_siglip_gap_378(pretrained: bool = False, **kwargs) -> VisionTransformer: def vit_so400m_patch14_siglip_gap_378(pretrained: bool = False, **kwargs) -> VisionTransformer:
""" A SigLIP variant of ViT with global average pooling (GAP) instead of attention pooling (MAP).""" """ A SigLIP variant of ViT with global average pooling (GAP) instead of attention pooling (MAP)."""
@ -3561,9 +3549,10 @@ def vit_so400m_patch14_siglip_gap_896(pretrained: bool = False, **kwargs) -> Vis
@register_model @register_model
def vit_so400m_patch16_siglip_gap_256(pretrained: bool = False, **kwargs) -> VisionTransformer: def vit_so400m_patch16_siglip_gap_256(pretrained: bool = False, **kwargs) -> VisionTransformer:
""" A SigLIP variant of ViT with global average pooling (GAP) instead of attention pooling (MAP)."""
model_args = dict( model_args = dict(
patch_size=16, embed_dim=1152, depth=27, num_heads=16, mlp_ratio=3.7362, class_token=False, patch_size=16, embed_dim=1152, depth=27, num_heads=16, mlp_ratio=3.7362,
global_pool='avg', fc_norm=False, act_layer='gelu_tanh' class_token=False, global_pool='avg', fc_norm=False, act_layer='gelu_tanh',
) )
model = _create_vision_transformer( model = _create_vision_transformer(
'vit_so400m_patch16_siglip_gap_256', pretrained=pretrained, **dict(model_args, **kwargs)) 'vit_so400m_patch16_siglip_gap_256', pretrained=pretrained, **dict(model_args, **kwargs))