Skip to content

A few more weights #2413

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jan 18, 2025
Merged

A few more weights #2413

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6b9c17f
Add some so150m vit w/ sbb recipe weights, and a ese_vovnet57b model …
rwightman Jan 17, 2025
c5cf0e0
Add the 256x256 in1k ft of the so150m, add an alternate so150m def
rwightman Jan 18, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 38 additions & 2 deletions timm/models/vision_transformer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2152,10 +2152,20 @@ def _cfg(url: str = '', **kwargs) -> Dict[str, Any]:
'vit_base_patch16_reg4_gap_256.untrained': _cfg(
input_size=(3, 256, 256)),

'vit_so150m_patch16_reg4_gap_256.untrained': _cfg(
input_size=(3, 256, 256)),
'vit_so150m_patch16_reg4_gap_256.sbb_e250_in12k_ft_in1k': _cfg(
hf_hub_id='timm/',
input_size=(3, 256, 256), crop_pct=0.95),
'vit_so150m_patch16_reg4_gap_256.sbb_e250_in12k': _cfg(
hf_hub_id='timm/',
num_classes=11821,
input_size=(3, 256, 256), crop_pct=0.95),
'vit_so150m_patch16_reg4_gap_384.sbb_e250_in12k_ft_in1k': _cfg(
hf_hub_id='timm/',
input_size=(3, 384, 384), crop_pct=1.0),
'vit_so150m_patch16_reg4_map_256.untrained': _cfg(
input_size=(3, 256, 256)),
'vit_so150m2_patch16_reg1_gap_256.untrained': _cfg(
input_size=(3, 256, 256), crop_pct=0.95),

'vit_intern300m_patch14_448.ogvl_dist': _cfg(
hf_hub_id='timm/',
Expand Down Expand Up @@ -3462,6 +3472,7 @@ def vit_base_patch16_reg4_gap_256(pretrained: bool = False, **kwargs) -> VisionT

@register_model
def vit_so150m_patch16_reg4_map_256(pretrained: bool = False, **kwargs) -> VisionTransformer:
""" SO150M (shape optimized, but diff than paper def, optimized for GPU) """
model_args = dict(
patch_size=16, embed_dim=896, depth=18, num_heads=14, mlp_ratio=2.572,
class_token=False, reg_tokens=4, global_pool='map',
Expand All @@ -3473,6 +3484,7 @@ def vit_so150m_patch16_reg4_map_256(pretrained: bool = False, **kwargs) -> Visio

@register_model
def vit_so150m_patch16_reg4_gap_256(pretrained: bool = False, **kwargs) -> VisionTransformer:
""" SO150M (shape optimized, but diff than paper def, optimized for GPU) """
model_args = dict(
patch_size=16, embed_dim=896, depth=18, num_heads=14, mlp_ratio=2.572,
class_token=False, reg_tokens=4, global_pool='avg', fc_norm=False,
Expand All @@ -3482,6 +3494,30 @@ def vit_so150m_patch16_reg4_gap_256(pretrained: bool = False, **kwargs) -> Visio
return model


@register_model
def vit_so150m_patch16_reg4_gap_384(pretrained: bool = False, **kwargs) -> VisionTransformer:
""" SO150M (shape optimized, but diff than paper def, optimized for GPU) """
model_args = dict(
patch_size=16, embed_dim=896, depth=18, num_heads=14, mlp_ratio=2.572,
class_token=False, reg_tokens=4, global_pool='avg', fc_norm=False,
)
model = _create_vision_transformer(
'vit_so150m_patch16_reg4_gap_384', pretrained=pretrained, **dict(model_args, **kwargs))
return model


@register_model
def vit_so150m2_patch16_reg1_gap_256(pretrained: bool = False, **kwargs) -> VisionTransformer:
""" SO150M v2 (shape optimized, but diff than paper def, optimized for GPU) """
model_args = dict(
patch_size=16, embed_dim=896, depth=20, num_heads=14, mlp_ratio=2.429, init_values=1e-5,
qkv_bias=False, class_token=False, reg_tokens=1, global_pool='avg',
)
model = _create_vision_transformer(
'vit_so150m2_patch16_reg1_gap_256', pretrained=pretrained, **dict(model_args, **kwargs))
return model


@register_model
def vit_intern300m_patch14_448(pretrained: bool = False, **kwargs) -> VisionTransformer:
model_args = dict(
Expand Down
7 changes: 6 additions & 1 deletion timm/models/vovnet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -419,7 +419,12 @@ def _cfg(url='', **kwargs):
'ese_vovnet39b.ra_in1k': _cfg(
hf_hub_id='timm/',
test_input_size=(3, 288, 288), test_crop_pct=0.95),
'ese_vovnet57b.untrained': _cfg(url=''),
'ese_vovnet57b.ra4_e3600_r256_in1k': _cfg(
hf_hub_id='timm/',
mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5),
crop_pct=0.95, input_size=(3, 256, 256), pool_size=(8, 8),
test_input_size=(3, 320, 320), test_crop_pct=1.0
),
'ese_vovnet99b.untrained': _cfg(url=''),
'eca_vovnet39b.untrained': _cfg(url=''),
'ese_vovnet39b_evos.untrained': _cfg(url=''),
Expand Down
Loading