Skip to content

Commit 8ece53e

Browse files
committed
Switch BEiT to HF hub weights
1 parent d1bfa9a commit 8ece53e

File tree

2 files changed

+28
-12
lines changed

2 files changed

+28
-12
lines changed

timm/models/beit.py

Lines changed: 24 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -355,64 +355,76 @@ def _cfg(url='', **kwargs):
355355

356356
default_cfgs = generate_default_cfgs({
357357
'beit_base_patch16_224.in22k_ft_in22k_in1k': _cfg(
358-
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_base_patch16_224_pt22k_ft22kto1k.pth'),
358+
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_base_patch16_224_pt22k_ft22kto1k.pth',
359+
hf_hub_id='timm/'),
359360
'beit_base_patch16_384.in22k_ft_in22k_in1k': _cfg(
360361
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_base_patch16_384_pt22k_ft22kto1k.pth',
362+
hf_hub_id='timm/',
361363
input_size=(3, 384, 384), crop_pct=1.0,
362364
),
363365
'beit_base_patch16_224.in22k_ft_in22k': _cfg(
364366
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_base_patch16_224_pt22k_ft22k.pth',
367+
hf_hub_id='timm/',
365368
num_classes=21841,
366369
),
367370
'beit_large_patch16_224.in22k_ft_in22k_in1k': _cfg(
368-
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_large_patch16_224_pt22k_ft22kto1k.pth'),
371+
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_large_patch16_224_pt22k_ft22kto1k.pth',
372+
hf_hub_id='timm/'),
369373
'beit_large_patch16_384.in22k_ft_in22k_in1k': _cfg(
370374
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_large_patch16_384_pt22k_ft22kto1k.pth',
375+
hf_hub_id='timm/',
371376
input_size=(3, 384, 384), crop_pct=1.0,
372377
),
373378
'beit_large_patch16_512.in22k_ft_in22k_in1k': _cfg(
374379
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_large_patch16_512_pt22k_ft22kto1k.pth',
380+
hf_hub_id='timm/',
375381
input_size=(3, 512, 512), crop_pct=1.0,
376382
),
377383
'beit_large_patch16_224.in22k_ft_in22k': _cfg(
378384
url='https://conversationhub.blob.core.windows.net/beit-share-public/beit/beit_large_patch16_224_pt22k_ft22k.pth',
385+
hf_hub_id='timm/',
379386
num_classes=21841,
380387
),
381388

382389
'beitv2_base_patch16_224.in1k_ft_in22k_in1k': _cfg(
383390
url='https://conversationhub.blob.core.windows.net/beit-share-public/beitv2/beitv2_base_patch16_224_pt1k_ft21kto1k.pth',
391+
hf_hub_id='timm/',
384392
mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD
385393
),
386394
'beitv2_base_patch16_224.in1k_ft_in22k': _cfg(
387395
url='https://conversationhub.blob.core.windows.net/beit-share-public/beitv2/beitv2_base_patch16_224_pt1k_ft21k.pth',
388-
num_classes=21841,
389-
mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD
396+
hf_hub_id='timm/',
397+
num_classes=21841, mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD
390398
),
391399
'beitv2_large_patch16_224.in1k_ft_in22k_in1k': _cfg(
392400
url='https://conversationhub.blob.core.windows.net/beit-share-public/beitv2/beitv2_large_patch16_224_pt1k_ft21kto1k.pth',
393-
crop_pct=0.95,
394-
mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD
401+
hf_hub_id='timm/',
402+
crop_pct=0.95, mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD
395403
),
396404
'beitv2_large_patch16_224.in1k_ft_in22k': _cfg(
397405
url='https://conversationhub.blob.core.windows.net/beit-share-public/beitv2/beitv2_large_patch16_224_pt1k_ft21k.pth',
398-
num_classes=21841,
399-
mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD
406+
hf_hub_id='timm/',
407+
num_classes=21841, mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD
400408
),
401409

402410
'eva_giant_patch14_224.clip_ft_in1k': _cfg(
403-
hf_hub_id='BAAI/EVA', hf_hub_filename='eva_clip_vis_enc_sz224_ftcls_89p1.pt',
411+
# hf_hub_id='BAAI/EVA', hf_hub_filename='eva_clip_vis_enc_sz224_ftcls_89p1.pt',
412+
hf_hub_id='timm/',
404413
mean=OPENAI_CLIP_MEAN, std=OPENAI_CLIP_STD, crop_pct=1.0,
405414
),
406415
'eva_giant_patch14_336.clip_ft_in1k': _cfg(
407-
hf_hub_id='BAAI/EVA', hf_hub_filename='eva_clip_vis_enc_sz336_ftcls_89p4.pt',
416+
# hf_hub_id='BAAI/EVA', hf_hub_filename='eva_clip_vis_enc_sz336_ftcls_89p4.pt',
417+
hf_hub_id='timm/',
408418
mean=OPENAI_CLIP_MEAN, std=OPENAI_CLIP_STD,
409419
input_size=(3, 336, 336), crop_pct=1.0, crop_mode='squash'),
410420
'eva_giant_patch14_336.m30m_ft_in22k_in1k': _cfg(
411-
hf_hub_id='BAAI/EVA', hf_hub_filename='eva_21k_1k_336px_psz14_ema_89p6.pt',
421+
# hf_hub_id='BAAI/EVA', hf_hub_filename='eva_21k_1k_336px_psz14_ema_89p6.pt',
422+
hf_hub_id='timm/',
412423
mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD,
413424
input_size=(3, 336, 336), crop_pct=1.0, crop_mode='squash'),
414425
'eva_giant_patch14_560.m30m_ft_in22k_in1k': _cfg(
415-
hf_hub_id='BAAI/EVA', hf_hub_filename='eva_21k_1k_560px_psz14_ema_89p7.pt',
426+
# hf_hub_id='BAAI/EVA', hf_hub_filename='eva_21k_1k_560px_psz14_ema_89p7.pt',
427+
hf_hub_id='timm/',
416428
mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD,
417429
input_size=(3, 560, 560), crop_pct=1.0, crop_mode='squash'),
418430
})

timm/models/vision_transformer.py

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1052,18 +1052,22 @@ def _cfg(url='', **kwargs):
10521052
# EVA fine-tuned weights from MAE style MIM - EVA-CLIP target pretrain
10531053
# https://github.com/baaivision/EVA/blob/7ecf2c0a370d97967e86d047d7af9188f78d2df3/eva/README.md#eva-l-learning-better-mim-representations-from-eva-clip
10541054
'eva_large_patch14_196.in22k_ft_in22k_in1k': _cfg(
1055+
# hf_hub_id='BAAI/EVA', hf_hub_filename='eva_l_psz14_196px_21k_to_1k_ft_88p6.pt',
10551056
hf_hub_id='timm/',
10561057
mean=OPENAI_CLIP_MEAN, std=OPENAI_CLIP_STD,
10571058
input_size=(3, 196, 196), crop_pct=1.0),
10581059
'eva_large_patch14_336.in22k_ft_in22k_in1k': _cfg(
1060+
# hf_hub_id='BAAI/EVA', hf_hub_filename='eva_l_psz14_336px_21k_to_1k_ft_89p2.pt',
10591061
hf_hub_id='timm/',
10601062
mean=OPENAI_CLIP_MEAN, std=OPENAI_CLIP_STD,
10611063
input_size=(3, 336, 336), crop_pct=1.0, crop_mode='squash'),
10621064
'eva_large_patch14_196.in22k_ft_in1k': _cfg(
1065+
# hf_hub_id='BAAI/EVA', hf_hub_filename='eva_l_psz14_196px_1k_ft_88p0.pt',
10631066
hf_hub_id='timm/',
10641067
mean=OPENAI_CLIP_MEAN, std=OPENAI_CLIP_STD,
10651068
input_size=(3, 196, 196), crop_pct=1.0),
10661069
'eva_large_patch14_336.in22k_ft_in1k': _cfg(
1070+
# hf_hub_id='BAAI/EVA', hf_hub_filename='eva_l_psz14_336px_1k_ft_88p65.pt',
10671071
hf_hub_id='timm/',
10681072
mean=OPENAI_CLIP_MEAN, std=OPENAI_CLIP_STD,
10691073
input_size=(3, 336, 336), crop_pct=1.0, crop_mode='squash'),

0 commit comments

Comments
 (0)