birder 0.3.3__py3-none-any.whl → 0.4.0__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (92) hide show
  1. birder/common/fs_ops.py +2 -2
  2. birder/introspection/attention_rollout.py +1 -1
  3. birder/introspection/transformer_attribution.py +1 -1
  4. birder/layers/layer_scale.py +1 -1
  5. birder/net/__init__.py +2 -10
  6. birder/net/_rope_vit_configs.py +430 -0
  7. birder/net/_vit_configs.py +479 -0
  8. birder/net/biformer.py +1 -0
  9. birder/net/cait.py +5 -5
  10. birder/net/coat.py +12 -12
  11. birder/net/conv2former.py +3 -3
  12. birder/net/convmixer.py +1 -1
  13. birder/net/convnext_v1.py +1 -1
  14. birder/net/crossvit.py +5 -5
  15. birder/net/davit.py +1 -1
  16. birder/net/deit.py +12 -26
  17. birder/net/deit3.py +42 -189
  18. birder/net/densenet.py +9 -8
  19. birder/net/detection/deformable_detr.py +5 -2
  20. birder/net/detection/detr.py +5 -2
  21. birder/net/detection/efficientdet.py +1 -1
  22. birder/net/dpn.py +1 -2
  23. birder/net/edgenext.py +2 -1
  24. birder/net/edgevit.py +3 -0
  25. birder/net/efficientformer_v1.py +2 -1
  26. birder/net/efficientformer_v2.py +18 -31
  27. birder/net/efficientnet_v2.py +3 -0
  28. birder/net/efficientvit_mit.py +5 -5
  29. birder/net/fasternet.py +2 -2
  30. birder/net/flexivit.py +22 -43
  31. birder/net/groupmixformer.py +1 -1
  32. birder/net/hgnet_v1.py +5 -5
  33. birder/net/inception_next.py +1 -1
  34. birder/net/inception_resnet_v1.py +3 -3
  35. birder/net/inception_resnet_v2.py +7 -4
  36. birder/net/inception_v3.py +3 -0
  37. birder/net/inception_v4.py +3 -0
  38. birder/net/maxvit.py +1 -1
  39. birder/net/metaformer.py +3 -3
  40. birder/net/mim/crossmae.py +1 -1
  41. birder/net/mim/mae_vit.py +1 -1
  42. birder/net/mim/simmim.py +1 -1
  43. birder/net/mobilenet_v1.py +0 -9
  44. birder/net/mobilenet_v2.py +38 -44
  45. birder/net/{mobilenet_v3_large.py → mobilenet_v3.py} +37 -10
  46. birder/net/mobilevit_v1.py +5 -32
  47. birder/net/mobilevit_v2.py +1 -45
  48. birder/net/moganet.py +8 -5
  49. birder/net/mvit_v2.py +6 -6
  50. birder/net/nfnet.py +4 -0
  51. birder/net/pit.py +1 -1
  52. birder/net/pvt_v1.py +5 -5
  53. birder/net/pvt_v2.py +5 -5
  54. birder/net/repghost.py +1 -30
  55. birder/net/resmlp.py +2 -2
  56. birder/net/resnest.py +3 -0
  57. birder/net/resnet_v1.py +125 -1
  58. birder/net/resnet_v2.py +75 -1
  59. birder/net/resnext.py +35 -1
  60. birder/net/rope_deit3.py +33 -136
  61. birder/net/rope_flexivit.py +18 -18
  62. birder/net/rope_vit.py +3 -735
  63. birder/net/simple_vit.py +22 -16
  64. birder/net/smt.py +1 -1
  65. birder/net/squeezenet.py +5 -12
  66. birder/net/squeezenext.py +0 -24
  67. birder/net/ssl/capi.py +1 -1
  68. birder/net/ssl/data2vec.py +1 -1
  69. birder/net/ssl/dino_v2.py +2 -2
  70. birder/net/ssl/franca.py +2 -2
  71. birder/net/ssl/i_jepa.py +1 -1
  72. birder/net/ssl/ibot.py +1 -1
  73. birder/net/swiftformer.py +12 -2
  74. birder/net/swin_transformer_v2.py +1 -1
  75. birder/net/tiny_vit.py +3 -16
  76. birder/net/van.py +2 -2
  77. birder/net/vit.py +35 -963
  78. birder/net/vit_sam.py +13 -38
  79. birder/net/xcit.py +7 -6
  80. birder/tools/introspection.py +1 -1
  81. birder/tools/model_info.py +3 -1
  82. birder/version.py +1 -1
  83. {birder-0.3.3.dist-info → birder-0.4.0.dist-info}/METADATA +1 -1
  84. {birder-0.3.3.dist-info → birder-0.4.0.dist-info}/RECORD +88 -90
  85. birder/net/mobilenet_v3_small.py +0 -43
  86. birder/net/se_resnet_v1.py +0 -105
  87. birder/net/se_resnet_v2.py +0 -59
  88. birder/net/se_resnext.py +0 -30
  89. {birder-0.3.3.dist-info → birder-0.4.0.dist-info}/WHEEL +0 -0
  90. {birder-0.3.3.dist-info → birder-0.4.0.dist-info}/entry_points.txt +0 -0
  91. {birder-0.3.3.dist-info → birder-0.4.0.dist-info}/licenses/LICENSE +0 -0
  92. {birder-0.3.3.dist-info → birder-0.4.0.dist-info}/top_level.txt +0 -0
birder/net/rope_deit3.py CHANGED
@@ -23,6 +23,11 @@ from torch import nn
23
23
 
24
24
  from birder.common.masking import mask_tensor
25
25
  from birder.model_registry import registry
26
+ from birder.net._vit_configs import BASE
27
+ from birder.net._vit_configs import LARGE
28
+ from birder.net._vit_configs import MEDIUM
29
+ from birder.net._vit_configs import SMALL
30
+ from birder.net._vit_configs import TINY
26
31
  from birder.net.base import DetectorBackbone
27
32
  from birder.net.base import MaskedTokenOmissionMixin
28
33
  from birder.net.base import MaskedTokenRetentionMixin
@@ -462,204 +467,96 @@ class RoPE_DeiT3(DetectorBackbone, PreTrainEncoder, MaskedTokenOmissionMixin, Ma
462
467
  registry.register_model_config(
463
468
  "rope_deit3_t16",
464
469
  RoPE_DeiT3,
465
- config={
466
- "patch_size": 16,
467
- "num_layers": 12,
468
- "num_heads": 3,
469
- "hidden_dim": 192,
470
- "mlp_dim": 768,
471
- "drop_path_rate": 0.0,
472
- },
470
+ config={"patch_size": 16, **TINY},
473
471
  )
474
472
  registry.register_model_config(
475
473
  "rope_deit3_s16",
476
474
  RoPE_DeiT3,
477
- config={
478
- "patch_size": 16,
479
- "num_layers": 12,
480
- "num_heads": 6,
481
- "hidden_dim": 384,
482
- "mlp_dim": 1536,
483
- "drop_path_rate": 0.05,
484
- },
475
+ config={"patch_size": 16, **SMALL, "drop_path_rate": 0.05},
485
476
  )
486
477
  registry.register_model_config(
487
478
  "rope_deit3_s14",
488
479
  RoPE_DeiT3,
489
- config={
490
- "patch_size": 14,
491
- "num_layers": 12,
492
- "num_heads": 6,
493
- "hidden_dim": 384,
494
- "mlp_dim": 1536,
495
- "drop_path_rate": 0.05,
496
- },
480
+ config={"patch_size": 14, **SMALL, "drop_path_rate": 0.05},
497
481
  )
498
482
  registry.register_model_config(
499
483
  "rope_deit3_m16",
500
484
  RoPE_DeiT3,
501
- config={
502
- "patch_size": 16,
503
- "num_layers": 12,
504
- "num_heads": 8,
505
- "hidden_dim": 512,
506
- "mlp_dim": 2048,
507
- "drop_path_rate": 0.1,
508
- },
485
+ config={"patch_size": 16, **MEDIUM, "drop_path_rate": 0.1},
509
486
  )
510
487
  registry.register_model_config(
511
488
  "rope_deit3_m14",
512
489
  RoPE_DeiT3,
513
- config={
514
- "patch_size": 14,
515
- "num_layers": 12,
516
- "num_heads": 8,
517
- "hidden_dim": 512,
518
- "mlp_dim": 2048,
519
- "drop_path_rate": 0.1,
520
- },
490
+ config={"patch_size": 14, **MEDIUM, "drop_path_rate": 0.1},
521
491
  )
522
492
  registry.register_model_config(
523
493
  "rope_deit3_b16",
524
494
  RoPE_DeiT3,
525
- config={
526
- "patch_size": 16,
527
- "num_layers": 12,
528
- "num_heads": 12,
529
- "hidden_dim": 768,
530
- "mlp_dim": 3072,
531
- "drop_path_rate": 0.2,
532
- },
495
+ config={"patch_size": 16, **BASE, "drop_path_rate": 0.2},
533
496
  )
534
497
  registry.register_model_config(
535
498
  "rope_deit3_b14",
536
499
  RoPE_DeiT3,
537
- config={
538
- "patch_size": 14,
539
- "num_layers": 12,
540
- "num_heads": 12,
541
- "hidden_dim": 768,
542
- "mlp_dim": 3072,
543
- "drop_path_rate": 0.2,
544
- },
500
+ config={"patch_size": 14, **BASE, "drop_path_rate": 0.2},
545
501
  )
546
502
  registry.register_model_config(
547
503
  "rope_deit3_l16",
548
504
  RoPE_DeiT3,
549
- config={
550
- "patch_size": 16,
551
- "num_layers": 24,
552
- "num_heads": 16,
553
- "hidden_dim": 1024,
554
- "mlp_dim": 4096,
555
- "drop_path_rate": 0.45,
556
- },
505
+ config={"patch_size": 16, **LARGE, "drop_path_rate": 0.45},
506
+ )
507
+ registry.register_model_config(
508
+ "rope_deit3_l14",
509
+ RoPE_DeiT3,
510
+ config={"patch_size": 14, **LARGE, "drop_path_rate": 0.45},
557
511
  )
558
512
 
559
513
  # With registers
514
+ ####################
515
+
560
516
  registry.register_model_config(
561
517
  "rope_deit3_reg4_t16",
562
518
  RoPE_DeiT3,
563
- config={
564
- "patch_size": 16,
565
- "num_layers": 12,
566
- "num_heads": 3,
567
- "hidden_dim": 192,
568
- "mlp_dim": 768,
569
- "num_reg_tokens": 4,
570
- "drop_path_rate": 0.0,
571
- },
519
+ config={"patch_size": 16, **TINY, "num_reg_tokens": 4},
572
520
  )
573
521
  registry.register_model_config(
574
522
  "rope_deit3_reg4_s16",
575
523
  RoPE_DeiT3,
576
- config={
577
- "patch_size": 16,
578
- "num_layers": 12,
579
- "num_heads": 6,
580
- "hidden_dim": 384,
581
- "mlp_dim": 1536,
582
- "num_reg_tokens": 4,
583
- "drop_path_rate": 0.05,
584
- },
524
+ config={"patch_size": 16, **SMALL, "num_reg_tokens": 4, "drop_path_rate": 0.05},
585
525
  )
586
526
  registry.register_model_config(
587
527
  "rope_deit3_reg4_s14",
588
528
  RoPE_DeiT3,
589
- config={
590
- "patch_size": 14,
591
- "num_layers": 12,
592
- "num_heads": 6,
593
- "hidden_dim": 384,
594
- "mlp_dim": 1536,
595
- "num_reg_tokens": 4,
596
- "drop_path_rate": 0.05,
597
- },
529
+ config={"patch_size": 14, **SMALL, "num_reg_tokens": 4, "drop_path_rate": 0.05},
598
530
  )
599
531
  registry.register_model_config(
600
532
  "rope_deit3_reg4_m16",
601
533
  RoPE_DeiT3,
602
- config={
603
- "patch_size": 16,
604
- "num_layers": 12,
605
- "num_heads": 8,
606
- "hidden_dim": 512,
607
- "mlp_dim": 2048,
608
- "num_reg_tokens": 4,
609
- "drop_path_rate": 0.1,
610
- },
534
+ config={"patch_size": 16, **MEDIUM, "num_reg_tokens": 4, "drop_path_rate": 0.1},
611
535
  )
612
536
  registry.register_model_config(
613
537
  "rope_deit3_reg4_m14",
614
538
  RoPE_DeiT3,
615
- config={
616
- "patch_size": 14,
617
- "num_layers": 12,
618
- "num_heads": 8,
619
- "hidden_dim": 512,
620
- "mlp_dim": 2048,
621
- "num_reg_tokens": 4,
622
- "drop_path_rate": 0.1,
623
- },
539
+ config={"patch_size": 14, **MEDIUM, "num_reg_tokens": 4, "drop_path_rate": 0.1},
624
540
  )
625
541
  registry.register_model_config(
626
542
  "rope_deit3_reg4_b16",
627
543
  RoPE_DeiT3,
628
- config={
629
- "patch_size": 16,
630
- "num_layers": 12,
631
- "num_heads": 12,
632
- "hidden_dim": 768,
633
- "mlp_dim": 3072,
634
- "num_reg_tokens": 4,
635
- "drop_path_rate": 0.2,
636
- },
544
+ config={"patch_size": 16, **BASE, "num_reg_tokens": 4, "drop_path_rate": 0.2},
637
545
  )
638
546
  registry.register_model_config(
639
547
  "rope_deit3_reg4_b14",
640
548
  RoPE_DeiT3,
641
- config={
642
- "patch_size": 14,
643
- "num_layers": 12,
644
- "num_heads": 12,
645
- "hidden_dim": 768,
646
- "mlp_dim": 3072,
647
- "num_reg_tokens": 4,
648
- "drop_path_rate": 0.2,
649
- },
549
+ config={"patch_size": 14, **BASE, "num_reg_tokens": 4, "drop_path_rate": 0.2},
650
550
  )
651
551
  registry.register_model_config(
652
552
  "rope_deit3_reg4_l16",
653
553
  RoPE_DeiT3,
654
- config={
655
- "patch_size": 16,
656
- "num_layers": 24,
657
- "num_heads": 16,
658
- "hidden_dim": 1024,
659
- "mlp_dim": 4096,
660
- "num_reg_tokens": 4,
661
- "drop_path_rate": 0.45,
662
- },
554
+ config={"patch_size": 16, **LARGE, "num_reg_tokens": 4, "drop_path_rate": 0.45},
555
+ )
556
+ registry.register_model_config(
557
+ "rope_deit3_reg4_l14",
558
+ RoPE_DeiT3,
559
+ config={"patch_size": 14, **LARGE, "num_reg_tokens": 4, "drop_path_rate": 0.45},
663
560
  )
664
561
 
665
562
  registry.register_weights(
birder/net/rope_flexivit.py CHANGED
@@ -21,6 +21,8 @@ from birder.layers import MultiHeadAttentionPool
21
21
  from birder.layers import SwiGLU_FFN
22
22
  from birder.layers.activations import get_activation_module
23
23
  from birder.model_registry import registry
24
+ from birder.net._vit_configs import BASE
25
+ from birder.net._vit_configs import SMALL
24
26
  from birder.net.base import DetectorBackbone
25
27
  from birder.net.base import MaskedTokenOmissionMixin
26
28
  from birder.net.base import MaskedTokenRetentionMixin
@@ -661,26 +663,24 @@ class RoPE_FlexiViT(DetectorBackbone, PreTrainEncoder, MaskedTokenOmissionMixin,
661
663
  registry.register_model_config(
662
664
  "rope_flexivit_s16",
663
665
  RoPE_FlexiViT,
664
- config={
665
- "patch_size": 16,
666
- "num_layers": 12,
667
- "num_heads": 6,
668
- "hidden_dim": 384,
669
- "mlp_dim": 1536,
670
- "drop_path_rate": 0.0,
671
- },
666
+ config={"patch_size": 16, **SMALL},
667
+ )
668
+ registry.register_model_config(
669
+ "rope_flexivit_b16",
670
+ RoPE_FlexiViT,
671
+ config={"patch_size": 16, **BASE},
672
+ )
673
+
674
+ # With registers
675
+ ####################
676
+
677
+ registry.register_model_config(
678
+ "rope_flexivit_reg1_s16",
679
+ RoPE_FlexiViT,
680
+ config={"patch_size": 16, **SMALL, "num_reg_tokens": 1},
672
681
  )
673
682
  registry.register_model_config(
674
683
  "rope_flexivit_reg4_b16_avg",
675
684
  RoPE_FlexiViT,
676
- config={
677
- "patch_size": 16,
678
- "num_layers": 12,
679
- "num_heads": 12,
680
- "hidden_dim": 768,
681
- "mlp_dim": 3072,
682
- "num_reg_tokens": 4,
683
- "class_token": False,
684
- "drop_path_rate": 0.1,
685
- },
685
+ config={"patch_size": 16, **BASE, "num_reg_tokens": 4, "class_token": False},
686
686
  )