MCPcopy Create free account

hub / github.com/NVlabs/RADIO / functions

Functions572 in github.com/NVlabs/RADIO

⨍ Functions 572◇ Types & classes 107
↓ 71 callersMethodto
(self, *args, **kwargs)
radio/adaptor_base.py:28
↓ 49 callersFunctionrank_print
(*args, **kwargs)
examples/common/utils.py:69
↓ 25 callersMethodeval
(self)
examples/position_predictor.py:374
↓ 15 callersMethod__init__
(self, c1, c2, n=1, shortcut=False, g=1, e=0.5, drop_path=None)
radio/eradio_model.py:44
↓ 13 callersFunctionload_model
(version: str, adaptor_names: str = None, use_huggingface: bool = False, use_local_lib: bool = True,
examples/common/model_loader.py:339
↓ 12 callersFunction_create_vision_transformer
(*args, **kwargs)
radio/extra_timm_models.py:164
↓ 12 callersMethodforward_intermediates
Forward features that returns intermediates. Args: x: Input image tensor indices: Take last n blocks if int, all if N
radio/radio1d.py:1518
↓ 8 callersMethod__init__
( self, in_features: int, hidden_features: Optional[int] = None, out_features:
radio/dinov2_arch.py:186
↓ 8 callersFunction_make_dinov2_model
( *, arch_name: str = "vit_large", img_size: int = 518, patch_size: int = 14, init_values:
radio/dinov2_arch.py:898
↓ 8 callersFunctionget_prefix_state_dict
(state_dict: Dict[str, Any], prefix: str)
hubconf.py:201
↓ 8 callersFunctionget_rank
(group: Optional[dist.ProcessGroup] = None)
radio/utils.py:6
↓ 6 callersMethod__init__
Args: img_size: Input image size. patch_size: Patch size. in_chans: Number of image input channels.
radio/radio1d.py:847
↓ 6 callersMethod__init__
(self, sam_encoder: nn.Module)
examples/common/model_loader.py:199
↓ 6 callersFunction_get_stats
(t: torch.Tensor)
examples/model_stats.py:160
↓ 6 callersMethodprepare_tokens_with_masks
(self, x, masks=None)
radio/dinov2_arch.py:716
↓ 5 callersFunctionbarrier
(group: Optional[dist.ProcessGroup] = None)
radio/utils.py:14
↓ 5 callersFunctionget_standard_transform
(resolution: List[int], resize_multiple: int, preprocessor = None, max_dim: bool = False, pad_mean = None)
examples/common/resize_transform.py:142
↓ 5 callersFunctionis_guidance_parametrized
(name: str)
radio/enable_spectral_reparam.py:153
↓ 5 callersFunctionradio_model
( version: str = "", progress: bool = True, adaptor_names: Union[str, List[str]] = None, neck_
hubconf.py:30
↓ 5 callersFunctionrank_print
(*args, **kwargs)
examples/position_predictor.py:119
↓ 5 callersFunctionround_up
(value, multiple: int)
examples/common/utils.py:13
↓ 4 callersFunctiondisable_spectral_reparam
(model: nn.Module)
radio/enable_spectral_reparam.py:211
↓ 4 callersFunctionget_pca_map
feature_map: (1, h, w, C) is the feature map of a single image.
examples/visualize_features.py:597
↓ 4 callersMethodset_curr_step
(self, step: int)
radio/radio1d.py:678
↓ 3 callersMethod__init__
( self, head_dim: int, base=50000, scale_base=512 )
radio/vision_transformer_xpos.py:49
↓ 3 callersMethod_apply_downscale
Apply patch merging downscale operation. Args: x: Input tensor of shape (B, N, C) where N = num_prefix_tokens + H*W d
radio/radio1d.py:1159
↓ 3 callersFunction_create_vision_transformer
(name, pretrained=False, **kwargs)
radio/radio1d.py:1769
↓ 3 callersMethod_extract_final
(self, x: torch.Tensor, y: torch.Tensor, feature_fmt: str = 'NLC')
radio/radio_model.py:211
↓ 3 callersFunction_gather_cat
(t)
examples/ssl_metrics.py:197
↓ 3 callersMethod_wrap_output
(self, token, features)
examples/common/model_loader.py:101
↓ 3 callersFunctionadd_text_overlay
Add text overlay to an image. Args: image: numpy array (H, W, C) in BGR format with values 0-255 text: Text to display
examples/visualize_features.py:481
↓ 3 callersMethodblocks
(self)
radio/radio_model.py:130
↓ 3 callersFunctioncreate_mlp_from_config
(version: str, input_dim: int, hidden_dim: int, output_dim: int, num_inner: int, is_summary: bool = True, **kw
radio/adaptor_module_factory.py:74
↓ 3 callersFunctionenable_spectral_reparam
(model: Union[nn.Module, List[nn.Module]], n_power_iterations: int = 1,
radio/enable_spectral_reparam.py:134
↓ 3 callersMethodforward_features
Forward pass.
examples/position_predictor.py:325
↓ 3 callersFunctionforward_intermediates
Forward features that returns intermediates. The Dense layer aggregation method is inspired from the paper: "Dense Connector for MLLMs" by Y
radio/forward_intermediates.py:30
↓ 3 callersMethodget_max_tokens
(self, outside_max: int)
radio/radio1d.py:682
↓ 3 callersFunctionget_mlp_info_from_state
(version: str, state: Dict[str, torch.Tensor], prefix: str = '', spectral_weights: bool = False)
radio/adaptor_module_factory.py:43
↓ 3 callersMethodinput_conditioner
(self)
radio/hf_model.py:171
↓ 3 callersFunctionparametrize_linear
(linear: nn.Linear)
radio/enable_spectral_reparam.py:161
↓ 3 callersMethodsample
(self, batch_size: int, max_tokens: int)
radio/radio1d.py:691
↓ 3 callersFunctionsave_animated_image
Save a list of numpy arrays as an animated image. Args: frames: List of numpy arrays (H, W, C) in BGR format with values 0-255
examples/visualize_features.py:535
↓ 3 callersFunctionsmooth_rank
(eig: torch.Tensor)
examples/ssl_metrics.py:513
↓ 3 callersFunctionwindow_partition
Function for partitioning image into windows and later do windowed attention Args: x: (B, C, H, W) window_size: window size
radio/eradio_model.py:144
↓ 2 callersMethod__init__
(self, embed_dim: int, dtype: torch.dtype = torch.float32)
radio/feature_normalizer.py:28
↓ 2 callersMethod__init__
(self, weight: torch.Tensor, *args, init_norm_to_current: bool = False, alpha: float = 0.05, version: int = 2,
radio/enable_spectral_reparam.py:27
↓ 2 callersMethod__init__
(self, patch_size: int)
radio/vit_patch_generator.py:277
↓ 2 callersMethod__init__
(self, src_dim: int, tgt_dim: int, gated: bool = True)
radio/dual_hybrid_vit.py:18
↓ 2 callersMethod__init__
(self, target_size: Union[int, Tuple[int, int]], pad_mean = None)
examples/common/resize_transform.py:116
↓ 2 callersFunction_add_norm
(n, p)
examples/model_stats.py:62
↓ 2 callersFunction_add_stats
(name: str, **kwargs)
examples/model_stats.py:175
↓ 2 callersFunction_all_to_all
(t: torch.Tensor)
examples/knn_classification.py:230
↓ 2 callersFunction_build_database
(dataset, model: nn.Module, device: torch.device, num_steps: int, rank: int, amp: bool = True, adaptor: str =
examples/knn_classification.py:342
↓ 2 callersFunction_dino_student
(arch: str, **kwargs)
radio/extra_models.py:184
↓ 2 callersFunction_enable_cpe_for_timm_vit
(model: VisionTransformer, forward_fn: Callable, max
radio/enable_cpe_support.py:90
↓ 2 callersFunction_get_dataset
(builder, split: str)
examples/knn_classification.py:139
↓ 2 callersFunction_get_dataset
(builder, split: str)
examples/position_predictor.py:186
↓ 2 callersFunction_get_paley_q
(q: int)
examples/common/hadamard.py:127
↓ 2 callersFunction_get_prime_factors
(n: int)
examples/common/hadamard.py:185
↓ 2 callersMethod_get_sigma
(self, weight: torch.Tensor, n_power_iterations: int = None)
radio/enable_spectral_reparam.py:56
↓ 2 callersFunction_get_warp
(features: torch.Tensor, transform: torch.Tensor, image_size)
examples/ssl_metrics.py:424
↓ 2 callersFunction_inner_replace_ln
(block: Block, depth: int, key: str)
radio/extra_timm_models.py:238
↓ 2 callersMethod_instantiate
(self, dist: Union[KSampleDistribution, dict, str])
radio/radio1d.py:812
↓ 2 callersFunction_is_paley_construction
(q: int, modulo: int)
examples/common/hadamard.py:159
↓ 2 callersMethod_patchify
(self, x: torch.Tensor)
radio/vision_transformer_xpos.py:319
↓ 2 callersFunction_prepare
(builder)
examples/position_predictor.py:164
↓ 2 callersFunction_replace_ln
(model: VisionTransformer, fn)
radio/extra_timm_models.py:237
↓ 2 callersFunction_round_up
(value, multiple: int)
examples/position_predictor.py:285
↓ 2 callersFunction_run_kernel
(x: torch.Tensor, mean: torch.Tensor, tx: torch.Tensor)
radio/feature_normalizer.py:14
↓ 2 callersFunction_save_image
(im: torch.Tensor, dir: str, path: str)
examples/ssl_metrics.py:586
↓ 2 callersFunction_to_tensor
(v: norm_t)
radio/input_conditioner.py:52
↓ 2 callersFunctionalbum_aug
(tx, image, bds)
examples/ssl_metrics.py:255
↓ 2 callersFunctioncalc_lidar_from_cov
(cov_b: torch.Tensor, cov_w: torch.Tensor, delta: float = DELTA)
examples/ssl_metrics.py:358
↓ 2 callersFunctioncalc_rank_me
(embeddings: torch.Tensor)
examples/ssl_metrics.py:319
↓ 2 callersFunctioncompute_smooth_rank
(features: torch.Tensor)
examples/cca_compare_models.py:164
↓ 2 callersFunctionconfigure_spectral_reparam_from_args
(model: nn.Module, args, state_dict_guidance: Optional[Dict[str, torch.Tensor]] = None)
radio/enable_spectral_reparam.py:197
↓ 2 callersFunctionconst_ste
Straight-through estimator that returns a constant `c` with the same shape as `x`, while routing gradients through `x` (identity).
radio/radio1d.py:221
↓ 2 callersFunctioncreate_image_grid_with_annotations
Arrange a nested list of MxN numpy arrays into a single image grid with column annotations. Args: images (list of list of np.array): Nes
examples/visualize_features.py:657
↓ 2 callersFunctioncreate_mlp_from_state
(version: str, state: Dict[str, torch.Tensor], prefix: str = '', spectral_weights: bool = False, is_summary: b
radio/adaptor_module_factory.py:82
↓ 2 callersFunctioncreate_model_from_args
(args)
radio/radio_model.py:344
↓ 2 callersFunctiondeterministic_grid_init
Return a diverse but deterministic grid of values.
test_hf.py:23
↓ 2 callersFunctiondrop_add_residual_stochastic_depth
( x: torch.Tensor, residual_func: Callable[[torch.Tensor], torch.Tensor], sample_drop_ratio: float
radio/dinov2_arch.py:448
↓ 2 callersFunctiondrop_add_residual_stochastic_depth_list
( x_list: List[torch.Tensor], residual_func: Callable[[torch.Tensor, Any], torch.Tensor], sample_d
radio/dinov2_arch.py:519
↓ 2 callersMethodencode_text
(self, text, normalize: bool = False, **kwargs)
examples/common/model_loader.py:119
↓ 2 callersMethodforward
(self, x: torch.Tensor)
radio/dinov2_arch.py:203
↓ 2 callersMethodforward_encoder
Forward pass through encoder only (embeddings, transformer blocks, token slicing). Args: x: Input image tensor of shape (B, C, H,
radio/radio1d.py:1198
↓ 2 callersMethodforward_features
Forward pass through encoder and decoder (full feature extraction). Args: x: Input image tensor of shape (B, C, H, W)
radio/radio1d.py:1479
↓ 2 callersFunctiongenerate_position_tensor
(predictions)
examples/position_predictor.py:378
↓ 2 callersFunctionget_a_to_b_fidelity
(a_features: torch.Tensor, b_features: torch.Tensor, proj_a: torch.Tensor, proj_b: torch.Tensor)
examples/cca_compare_models.py:155
↓ 2 callersFunctionget_attn_bias_and_cat
this will perform the index select, cat the tensors, and provide the attn_bias from cache
radio/dinov2_arch.py:495
↓ 2 callersFunctionget_bernoulli_matrix
(feature_dim: int)
examples/common/hadamard.py:148
↓ 2 callersFunctionget_default_conditioner
()
radio/input_conditioner.py:42
↓ 2 callersMethodget_expected_tokens
(self, outside_max: int)
radio/radio1d.py:685
↓ 2 callersFunctionget_feature_matrix
( cache_dir: str, model_name: str, which: str, adaptor: str, *args, **kwargs )
examples/cca_compare_models.py:136
↓ 2 callersFunctionget_phi_s_matrix
(X: torch.Tensor)
examples/common/phi_s.py:8
↓ 2 callersFunctionget_robust_pca
(features: torch.Tensor, m: float = 2, remove_first_component=False, skip: int = 0)
examples/visualize_features.py:562
↓ 2 callersFunctionget_sylvester_hadamard_matrix
(feature_dim: int)
examples/common/hadamard.py:62
↓ 2 callersFunctionget_warped_features
(features: torch.Tensor, transform: torch.Tensor, image_size)
examples/ssl_metrics.py:600
↓ 2 callersFunctionget_world_size
(group: Optional[dist.ProcessGroup] = None)
radio/utils.py:10
↓ 2 callersFunctioninvsqrt
(v: torch.Tensor)
examples/cca_compare_models.py:245
next →1–100 of 572, ranked by callers