PyPI - birder - Versions diffs - 0.3.3__py3-none-any.whl → 0.4.0__py3-none-any.whl - Mend

birder 0.3.3py3-none-any.whl → 0.4.0py3-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

birder/common/fs_ops.py +2 -2
birder/introspection/attention_rollout.py +1 -1
birder/introspection/transformer_attribution.py +1 -1
birder/layers/layer_scale.py +1 -1
birder/net/__init__.py +2 -10
birder/net/_rope_vit_configs.py +430 -0
birder/net/_vit_configs.py +479 -0
birder/net/biformer.py +1 -0
birder/net/cait.py +5 -5
birder/net/coat.py +12 -12
birder/net/conv2former.py +3 -3
birder/net/convmixer.py +1 -1
birder/net/convnext_v1.py +1 -1
birder/net/crossvit.py +5 -5
birder/net/davit.py +1 -1
birder/net/deit.py +12 -26
birder/net/deit3.py +42 -189
birder/net/densenet.py +9 -8
birder/net/detection/deformable_detr.py +5 -2
birder/net/detection/detr.py +5 -2
birder/net/detection/efficientdet.py +1 -1
birder/net/dpn.py +1 -2
birder/net/edgenext.py +2 -1
birder/net/edgevit.py +3 -0
birder/net/efficientformer_v1.py +2 -1
birder/net/efficientformer_v2.py +18 -31
birder/net/efficientnet_v2.py +3 -0
birder/net/efficientvit_mit.py +5 -5
birder/net/fasternet.py +2 -2
birder/net/flexivit.py +22 -43
birder/net/groupmixformer.py +1 -1
birder/net/hgnet_v1.py +5 -5
birder/net/inception_next.py +1 -1
birder/net/inception_resnet_v1.py +3 -3
birder/net/inception_resnet_v2.py +7 -4
birder/net/inception_v3.py +3 -0
birder/net/inception_v4.py +3 -0
birder/net/maxvit.py +1 -1
birder/net/metaformer.py +3 -3
birder/net/mim/crossmae.py +1 -1
birder/net/mim/mae_vit.py +1 -1
birder/net/mim/simmim.py +1 -1
birder/net/mobilenet_v1.py +0 -9
birder/net/mobilenet_v2.py +38 -44
birder/net/{mobilenet_v3_large.py → mobilenet_v3.py} +37 -10
birder/net/mobilevit_v1.py +5 -32
birder/net/mobilevit_v2.py +1 -45
birder/net/moganet.py +8 -5
birder/net/mvit_v2.py +6 -6
birder/net/nfnet.py +4 -0
birder/net/pit.py +1 -1
birder/net/pvt_v1.py +5 -5
birder/net/pvt_v2.py +5 -5
birder/net/repghost.py +1 -30
birder/net/resmlp.py +2 -2
birder/net/resnest.py +3 -0
birder/net/resnet_v1.py +125 -1
birder/net/resnet_v2.py +75 -1
birder/net/resnext.py +35 -1
birder/net/rope_deit3.py +33 -136
birder/net/rope_flexivit.py +18 -18
birder/net/rope_vit.py +3 -735
birder/net/simple_vit.py +22 -16
birder/net/smt.py +1 -1
birder/net/squeezenet.py +5 -12
birder/net/squeezenext.py +0 -24
birder/net/ssl/capi.py +1 -1
birder/net/ssl/data2vec.py +1 -1
birder/net/ssl/dino_v2.py +2 -2
birder/net/ssl/franca.py +2 -2
birder/net/ssl/i_jepa.py +1 -1
birder/net/ssl/ibot.py +1 -1
birder/net/swiftformer.py +12 -2
birder/net/swin_transformer_v2.py +1 -1
birder/net/tiny_vit.py +3 -16
birder/net/van.py +2 -2
birder/net/vit.py +35 -963
birder/net/vit_sam.py +13 -38
birder/net/xcit.py +7 -6
birder/tools/introspection.py +1 -1
birder/tools/model_info.py +3 -1
birder/version.py +1 -1
{birder-0.3.3.dist-info → birder-0.4.0.dist-info}/METADATA +1 -1
{birder-0.3.3.dist-info → birder-0.4.0.dist-info}/RECORD +88 -90
birder/net/mobilenet_v3_small.py +0 -43
birder/net/se_resnet_v1.py +0 -105
birder/net/se_resnet_v2.py +0 -59
birder/net/se_resnext.py +0 -30
{birder-0.3.3.dist-info → birder-0.4.0.dist-info}/WHEEL +0 -0
{birder-0.3.3.dist-info → birder-0.4.0.dist-info}/entry_points.txt +0 -0
{birder-0.3.3.dist-info → birder-0.4.0.dist-info}/licenses/LICENSE +0 -0
{birder-0.3.3.dist-info → birder-0.4.0.dist-info}/top_level.txt +0 -0

birder/net/rope_vit.py CHANGED Viewed

@@ -31,6 +31,7 @@ from birder.layers import MultiHeadAttentionPool
 from birder.layers import SwiGLU_FFN
 from birder.layers.activations import get_activation_module
 from birder.model_registry import registry
+from birder.net._rope_vit_configs import register_rope_vit_configs
 from birder.net.base import DetectorBackbone
 from birder.net.base import MaskedTokenOmissionMixin
 from birder.net.base import MaskedTokenRetentionMixin
@@ -939,741 +940,8 @@ class RoPE_ViT(DetectorBackbone, PreTrainEncoder, MaskedTokenOmissionMixin, Mask
         )
-# Vision Transformer Model Naming Convention
-# ==========================================
-#
-# Model names follow a structured pattern to encode architectural choices:
-# [rope_]vit_[reg{N}_][size][patch_size][_components][_pooling][_c{N}]
-#
-# Core Components:
-# - rope_       : Rotary Position Embedding (RoPE) enabled
-# - rope_i_     : Rotary Position Embedding (RoPE) enabled with interleaved rotation - implies different temp, indexing
-# - vit_        : Vision Transformer base architecture
-# - reg{N}_     : Register tokens (N = number of register tokens, e.g., reg4, reg8)
-# - size        : Model size (s=small, b=base, l=large, or specific like so150m)
-# - patch_size  : Patch size (e.g., 14, 16, 32 for 14x14, 16x16, 32x32 patches)
-#
-# Optional Components:
-#     Position Embeddings:
-#     - nps         : No Position embedding on Special tokens
-#
-#     Normalization:
-#     - rms         : RMSNorm (instead of LayerNorm)
-#     - pn          : Pre-Norm (layer norm before the encoder) - implies different norm eps
-#     - npn         : No Post Norm (disables post-normalization layer)
-#     - qkn         : QK Norm
-#
-#     Feed-Forward Network:
-#     - swiglu      : SwiGLU FFN layer type (instead of standard FFN)
-#
-#     Activation:
-#     - quick_gelu  : QuickGELU activation type
-#     - ...
-#
-#     Regularization:
-#     - ls          : Layer Scaling applied
-#
-#     Pooling/Reduction:
-#     - avg         : Average pooling for sequence reduction
-#     - ap          : Attention Pooling for sequence reduction
-#     - aps         : Attention Pooling inc. Special tokens for sequence reduction
-#
-#     Custom Variants:
-#     - c{N}        : Custom variant (N = version number) for models with fine-grained or non-standard
-#                     modifications not fully reflected in the name
-registry.register_model_config(
-    "rope_vit_s32",
-    RoPE_ViT,
-    config={
-        "patch_size": 32,
-        "num_layers": 12,
-        "num_heads": 6,
-        "hidden_dim": 384,
-        "mlp_dim": 1536,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_s16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 6,
-        "hidden_dim": 384,
-        "mlp_dim": 1536,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_i_vit_s16_pn_aps_c1",  # For PE Core - https://arxiv.org/abs/2504.13181
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 6,
-        "hidden_dim": 384,
-        "mlp_dim": 1536,
-        "pre_norm": True,
-        "attn_pool_head": True,
-        "attn_pool_num_heads": 8,
-        "attn_pool_special_tokens": True,
-        "norm_layer_eps": 1e-5,
-        "rope_rot_type": "interleaved",
-        "rope_grid_indexing": "xy",
-        "rope_grid_offset": 1,
-        "rope_temperature": 10000.0,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_s14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 12,
-        "num_heads": 6,
-        "hidden_dim": 384,
-        "mlp_dim": 1536,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_m32",
-    RoPE_ViT,
-    config={
-        "patch_size": 32,
-        "num_layers": 12,
-        "num_heads": 8,
-        "hidden_dim": 512,
-        "mlp_dim": 2048,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_m16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 8,
-        "hidden_dim": 512,
-        "mlp_dim": 2048,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_m14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 12,
-        "num_heads": 8,
-        "hidden_dim": 512,
-        "mlp_dim": 2048,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_b32",
-    RoPE_ViT,
-    config={
-        "patch_size": 32,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_b16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_b16_qkn_ls",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "layer_scale_init_value": 1e-5,
-        "qk_norm": True,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_i_vit_b16_pn_aps_c1",  # For PE Core - https://arxiv.org/abs/2504.13181
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "pre_norm": True,
-        "attn_pool_head": True,
-        "attn_pool_num_heads": 8,
-        "attn_pool_special_tokens": True,
-        "norm_layer_eps": 1e-5,
-        "rope_rot_type": "interleaved",
-        "rope_grid_indexing": "xy",
-        "rope_grid_offset": 1,
-        "rope_temperature": 10000.0,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_b14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_l32",
-    RoPE_ViT,
-    config={
-        "patch_size": 32,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_l16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_l14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_i_vit_l14_pn_aps_c1",  # For PE Core - https://arxiv.org/abs/2504.13181
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "pre_norm": True,
-        "attn_pool_head": True,
-        "attn_pool_num_heads": 8,
-        "attn_pool_special_tokens": True,
-        "norm_layer_eps": 1e-5,
-        "rope_rot_type": "interleaved",
-        "rope_grid_indexing": "xy",
-        "rope_grid_offset": 1,
-        "rope_temperature": 10000.0,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_h16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 32,
-        "num_heads": 16,
-        "hidden_dim": 1280,
-        "mlp_dim": 5120,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_h14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 32,
-        "num_heads": 16,
-        "hidden_dim": 1280,
-        "mlp_dim": 5120,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(  # From "Scaling Vision Transformers"
-    "rope_vit_g14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 40,
-        "num_heads": 16,
-        "hidden_dim": 1408,
-        "mlp_dim": 6144,
-        "drop_path_rate": 0.1,
-    },
-)
-# With registers
-registry.register_model_config(
-    "rope_vit_reg1_s32",
-    RoPE_ViT,
-    config={
-        "patch_size": 32,
-        "num_layers": 12,
-        "num_heads": 6,
-        "hidden_dim": 384,
-        "mlp_dim": 1536,
-        "num_reg_tokens": 1,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg1_s16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 6,
-        "hidden_dim": 384,
-        "mlp_dim": 1536,
-        "num_reg_tokens": 1,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_i_vit_reg1_s16_pn_npn_avg_c1",  # For PE Spatial - https://arxiv.org/abs/2504.13181
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 6,
-        "hidden_dim": 384,
-        "mlp_dim": 1536,
-        "num_reg_tokens": 1,
-        "class_token": False,
-        "pre_norm": True,
-        "post_norm": False,
-        "norm_layer_eps": 1e-5,
-        "rope_rot_type": "interleaved",
-        "rope_grid_indexing": "xy",
-        "rope_grid_offset": 1,
-        "rope_temperature": 10000.0,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg1_s14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 12,
-        "num_heads": 6,
-        "hidden_dim": 384,
-        "mlp_dim": 1536,
-        "num_reg_tokens": 1,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_m32",
-    RoPE_ViT,
-    config={
-        "patch_size": 32,
-        "num_layers": 12,
-        "num_heads": 8,
-        "hidden_dim": 512,
-        "mlp_dim": 2048,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_m16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 8,
-        "hidden_dim": 512,
-        "mlp_dim": 2048,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_m16_rms_avg",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 8,
-        "hidden_dim": 512,
-        "mlp_dim": 2048,
-        "num_reg_tokens": 4,
-        "class_token": False,
-        "norm_layer_type": "RMSNorm",
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_m14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 12,
-        "num_heads": 8,
-        "hidden_dim": 512,
-        "mlp_dim": 2048,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_m14_avg",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 12,
-        "num_heads": 8,
-        "hidden_dim": 512,
-        "mlp_dim": 2048,
-        "num_reg_tokens": 4,
-        "class_token": False,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_b32",
-    RoPE_ViT,
-    config={
-        "patch_size": 32,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.0,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_b16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_b14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg8_nps_b14_ap",
-    RoPE_ViT,
-    config={
-        "pos_embed_special_tokens": False,
-        "patch_size": 14,
-        "num_layers": 12,
-        "num_heads": 12,
-        "hidden_dim": 768,
-        "mlp_dim": 3072,
-        "num_reg_tokens": 8,
-        "class_token": False,
-        "attn_pool_head": True,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_l32",
-    RoPE_ViT,
-    config={
-        "patch_size": 32,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_l16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_l14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg8_l14_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "num_reg_tokens": 8,
-        "class_token": False,
-        "attn_pool_head": True,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg8_l14_rms_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 24,
-        "num_heads": 16,
-        "hidden_dim": 1024,
-        "mlp_dim": 4096,
-        "num_reg_tokens": 8,
-        "class_token": False,
-        "attn_pool_head": True,
-        "norm_layer_type": "RMSNorm",
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_h16",
-    RoPE_ViT,
-    config={
-        "patch_size": 16,
-        "num_layers": 32,
-        "num_heads": 16,
-        "hidden_dim": 1280,
-        "mlp_dim": 5120,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_h14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 32,
-        "num_heads": 16,
-        "hidden_dim": 1280,
-        "mlp_dim": 5120,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(  # From "Scaling Vision Transformers"
-    "rope_vit_reg4_g14",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 40,
-        "num_heads": 16,
-        "hidden_dim": 1408,
-        "mlp_dim": 6144,
-        "num_reg_tokens": 4,
-        "drop_path_rate": 0.1,
-    },
-)
-# Shape-optimized vision transformer (SoViT)
-registry.register_model_config(
-    "rope_vit_so150m_p14_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 18,
-        "num_heads": 16,
-        "hidden_dim": 896,  # Changed from 880 for RoPE divisibility
-        "mlp_dim": 2320,
-        "class_token": False,
-        "attn_pool_head": True,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_so400m_p14_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 27,
-        "num_heads": 16,
-        "hidden_dim": 1152,
-        "mlp_dim": 4304,
-        "class_token": False,
-        "attn_pool_head": True,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_so150m_p14_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 18,
-        "num_heads": 16,
-        "hidden_dim": 896,  # Changed from 880 for RoPE divisibility
-        "mlp_dim": 2320,
-        "num_reg_tokens": 4,
-        "class_token": False,
-        "attn_pool_head": True,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg8_so150m_p14_swiglu_rms_avg",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 18,
-        "num_heads": 16,
-        "hidden_dim": 896,  # Changed from 880 for RoPE divisibility
-        "mlp_dim": 2320,
-        "num_reg_tokens": 8,
-        "class_token": False,
-        "norm_layer_type": "RMSNorm",
-        "mlp_layer_type": "SwiGLU_FFN",
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg8_so150m_p14_swiglu_rms_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 18,
-        "num_heads": 16,
-        "hidden_dim": 896,  # Changed from 880 for RoPE divisibility
-        "mlp_dim": 2320,
-        "num_reg_tokens": 8,
-        "class_token": False,
-        "attn_pool_head": True,
-        "norm_layer_type": "RMSNorm",
-        "mlp_layer_type": "SwiGLU_FFN",
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg8_so150m_p14_swiglu_rms_aps",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 18,
-        "num_heads": 16,
-        "hidden_dim": 896,  # Changed from 880 for RoPE divisibility
-        "mlp_dim": 2320,
-        "num_reg_tokens": 8,
-        "class_token": False,
-        "attn_pool_head": True,
-        "attn_pool_special_tokens": True,
-        "norm_layer_type": "RMSNorm",
-        "mlp_layer_type": "SwiGLU_FFN",
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg8_so150m_p14_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 18,
-        "num_heads": 16,
-        "hidden_dim": 896,  # Changed from 880 for RoPE divisibility
-        "mlp_dim": 2320,
-        "num_reg_tokens": 8,
-        "class_token": False,
-        "attn_pool_head": True,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg4_so400m_p14_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 27,
-        "num_heads": 16,
-        "hidden_dim": 1152,
-        "mlp_dim": 4304,
-        "num_reg_tokens": 4,
-        "class_token": False,
-        "attn_pool_head": True,
-        "drop_path_rate": 0.1,
-    },
-)
-registry.register_model_config(
-    "rope_vit_reg8_so400m_p14_ap",
-    RoPE_ViT,
-    config={
-        "patch_size": 14,
-        "num_layers": 27,
-        "num_heads": 16,
-        "hidden_dim": 1152,
-        "mlp_dim": 4304,
-        "num_reg_tokens": 8,
-        "class_token": False,
-        "attn_pool_head": True,
-        "drop_path_rate": 0.1,
-    },
-)
+# Register model configs (side effects)
+register_rope_vit_configs(RoPE_ViT)
 registry.register_weights(
     "rope_vit_reg4_b14_capi",

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

birder 0.3.3py3-none-any.whl → 0.4.0py3-none-any.whl