Functions — github.com/UCSC-VLAA/OpenVision

hub / github.com/UCSC-VLAA/OpenVision / functions

Functions716 in github.com/UCSC-VLAA/OpenVision

⨍ Functions 716◇ Types & classes 90 ↳ Endpoints 1

↓ 2 callersFunction_process_batch

(batch_classnames)

src/convert_upload/open_clip/zero_shot_classifier.py:52

↓ 2 callersFunction_rescan_model_configs

()

src/convert_upload/open_clip/factory.py:33

↓ 2 callersMethod_sample_from_beta

(alpha, beta, shape)

src/transforms/mixup.py:112

↓ 2 callersFunction_split_frozen

Computes `frozen_mask` and updates `masks` and `scheds`.

src/optim/build_optax.py:287

↓ 2 callersFunction_token_to_tensor

(token_id, device: str = "cpu")

src/convert_upload/open_clip/coca_model.py:81

↓ 2 callersFunction_traverse_with_names

Traverses nested dicts/dataclasses and emits (leaf_name, leaf_val).

src/helpers/utils.py:445

↓ 2 callersFunction_window_mask

Mask a part of the image.

src/transforms/mixup.py:304

↓ 2 callersFunction_with_infinite_padding

Adds "infinite padding" to the dataset.

src/evaluators/proj/image_text/retrieval.py:57

↓ 2 callersFunction_with_infinite_padding

Adds "infinite padding" to the dataset.

src/evaluators/proj/image_text/discriminative_classifier.py:51

↓ 2 callersFunctionapply_transform

Apply the i-th transformation.

src/transforms/simclr_aug.py:97

↓ 2 callersMethodbatch_encode_plus

(self, texts, max_length=None)

src/convert_upload/open_clip/tokenizer.py:539

↓ 2 callersFunctionblockwise_ffn

(remat_ffn, inputs, chunk_size=2048, deterministic=True)

src/models/bpt.py:21

↓ 2 callersFunctionbuild_model_from_openai_state_dict

( state_dict: dict, quick_gelu=True, cast_dtype=torch.float16, )

src/convert_upload/open_clip/model.py:450

↓ 2 callersFunctionbytes_to_unicode

Returns list of utf-8 byte and a corresponding list of unicode strings. The reversible bpe codes work on unicode strings. This means you

src/convert_upload/open_clip/tokenizer.py:33

↓ 2 callersFunctioncanonicalize_text

Returns canonicalized `text` (lowercase and punctuation removed). From: https://github.com/google-research/big_vision/blob/53f18caf27a9419231bbf0

src/convert_upload/open_clip/tokenizer.py:105

↓ 2 callersFunctionconvert_mobile_clip_state_dict

(model: CustomTextCLIP, state_dict, fastvit = True)

src/convert_upload/open_clip/convert.py:142

↓ 2 callersFunctionconvert_weights_to_lp

Convert applicable model parameters to low-precision (bf16 or fp16)

src/convert_upload/open_clip/model.py:398

↓ 2 callersFunctioncreate_model

( model_name: str, pretrained: Optional[str] = None, precision: str = 'fp32',

src/convert_upload/open_clip/factory.py:196

↓ 2 callersFunctioncreate_model_from_pretrained

( model_name: str, pretrained: Optional[str] = None, precision: str = 'fp32',

src/convert_upload/open_clip/factory.py:453

↓ 2 callersFunctioncreate_vision_encoder_and_transforms

Creates a vision encoder model and its corresponding image transforms. This version correctly handles models from Hugging Face Hub.

src/convert_upload/open_clip/factory.py:502

↓ 2 callersFunctiondecode_variant

Converts a string like "B" or "B/32" into a params dict.

src/models/vit.py:816

↓ 2 callersMethoddist_loss

(self, teacher_logits, student_logits)

src/convert_upload/open_clip/loss.py:182

↓ 2 callersMethoddistort

Applies Mixup and/or Cutmix to batch of images and transforms labels. Args: images (tf.Tensor): Of shape [batch_size, height, width, 3] repr

src/transforms/mixup.py:77

↓ 2 callersFunctiondownload_pretrained

( cfg: Dict, force_hf_hub: bool = False, cache_dir: Union[str, None] = None, )

src/convert_upload/open_clip/pretrained.py:627

↓ 2 callersFunctiondownload_pretrained_from_url

( url: str, cache_dir: Union[str, None] = None, )

src/convert_upload/open_clip/pretrained.py:562

↓ 2 callersMethodencode_image

(self, image, normalize: bool = False)

src/convert_upload/open_clip/model.py:358

↓ 2 callersMethodencode_text

(self, text, normalize: bool = False)

src/convert_upload/open_clip/model.py:362

↓ 2 callersFunctionenqueue

(n_steps)

src/datasets/input_pipeline.py:199

↓ 2 callersFunctionget_1d_sincos_pos_embed_from_grid

embed_dim: output dimension for each position pos: a list of positions to be encoded: size (M,) out: (M, D)

src/convert_upload/open_clip/pos_embed.py:49

↓ 2 callersFunctionget_2d_sincos_pos_embed_from_grid

(embed_dim, grid)

src/convert_upload/open_clip/pos_embed.py:38

↓ 2 callersMethodget_cast_dtype

(self)

src/convert_upload/open_clip/transformer.py:350

↓ 2 callersMethodget_ground_truth

(self, device, num_logits)

src/convert_upload/open_clip/loss.py:89

↓ 2 callersFunctionget_pairs

Return set of symbol pairs in a word. Word is represented as tuple of symbols (symbols being variable-length strings).

src/convert_upload/open_clip/tokenizer.py:55

↓ 2 callersFunctionget_pretrained_url

(model: str, tag: str)

src/convert_upload/open_clip/pretrained.py:557

↓ 2 callersFunctionget_resize

Resizes image to a given size. Args: size: either an integer H, where H is both the new height and width of the resized image, or a list

src/transforms/ops_image.py:54

↓ 2 callersFunctionget_tokenizer

( model_name: str = '', context_length: Optional[int] = None, **kwargs, )

src/convert_upload/open_clip/factory.py:94

↓ 2 callersMethodget_transformer_kwargs

Helper to gather shared Transformer arguments.

src/models/text_decoder_v2.py:450

↓ 2 callersFunctionhas_hf_hub

(necessary=False)

src/convert_upload/open_clip/pretrained.py:608

↓ 2 callersFunctionlist_models

enumerate available model architectures based on config files

src/convert_upload/open_clip/factory.py:67

↓ 2 callersFunctionlist_pretrained_tags_by_model

return all pretrain tags for the specified model architecture

src/convert_upload/open_clip/pretrained.py:536

↓ 2 callersFunctionload_checkpoint

( model: Union[CLIP, CustomTextCLIP], checkpoint_path: str, strict: bool = True, )

src/convert_upload/open_clip/factory.py:159

↓ 2 callersFunctionload_params

Loads a parameters from a npz checkpoint. Args: tree: deprecated, use None. Bwd-compat for old format that only stored values: the pytree

src/helpers/utils.py:165

↓ 2 callersFunctionload_state_dict

(checkpoint_path: str, map_location='cpu')

src/convert_upload/open_clip/factory.py:144

↓ 2 callersMethodlookup

Lookup a name in the registry.

src/helpers/registry.py:154

↓ 2 callersFunctionmake_for_train

Makes an input pipeline for training.

src/datasets/input_pipeline.py:35

↓ 2 callersFunctionmaybe_cleanup_workdir

Potentially removes workdirs at end of run for cleanup.

src/helpers/utils.py:896

↓ 2 callersFunctionmaybe_pad

(x, actually_pad=True)

src/helpers/utils.py:94

↓ 2 callersFunctionneighbour_exchange_with_grad

(from_rank, to_rank, tensor, group=None)

src/convert_upload/open_clip/loss.py:285

↓ 2 callersFunctionposemb_sincos_2d

Follows the MoCo v3 logic.

src/models/vit.py:152

↓ 2 callersFunctionrandom_brightness

A multiplicative vs additive change of brightness.

src/transforms/simclr_aug.py:33

↓ 2 callersFunctionrecall_at

(k)

src/evaluators/proj/image_text/image_text_retrieval.py:47

↓ 2 callersMethodset_grad_checkpointing

(self, enable=True)

src/convert_upload/open_clip/model.py:263

↓ 2 callersFunctionsteps

Gets duration named `prefix` out of `config` and converts it to steps. Using this function to access a configuration value that denotes some kind

src/helpers/utils.py:925

↓ 2 callersFunctiontree_broadcast

Broadcasts a prefix tree to a full tree. Input-output examples: 1. prefix: {"x": 10, "y": 20} target: {"x": {"a": 1, "b": 2}, "y": 3}

src/helpers/utils.py:475

↓ 2 callersFunctiontree_map_with_names

Like jax.tree_util.tree_map but with a filter on the leaf path name. Args: f: A function with first parameter `name` (path-like "a/b/c") and re

src/helpers/utils.py:543

↓ 2 callersFunctionwhitespace_clean

(text)

src/convert_upload/open_clip/tokenizer.py:73

↓ 1 callersMethod__init__

(self, p=0.2)

src/convert_upload/open_clip/transform.py:262

↓ 1 callersMethod__init__

( self, embed_dim: int, vision_cfg: CLIPVisionCfg, text_cfg: C

src/convert_upload/open_clip/model.py:225

↓ 1 callersFunction_add_mask

(pp_fn)

src/datasets/input_pipeline.py:179

↓ 1 callersFunction_all_gather

All gather and flatten first two dims.

src/evaluators/proj/image_text/contrastive.py:41

↓ 1 callersFunction_apply_func_with_prob

Apply `func` to image w/ `args` as input with probability `prob`.

src/transforms/autoaugment.py:604

↓ 1 callersFunction_aspect75

(sz)

src/configs/clip_common.py:51

↓ 1 callersFunction_average_embeddings

Computes per-class averages of `embeddings`.

src/evaluators/proj/image_text/discriminative_classifier.py:156

↓ 1 callersFunction_build_text_decoder_tower

( embed_dim, multimodal_cfg, quick_gelu: bool = False, cast_dtype: Optional[to

src/convert_upload/open_clip/coca_model.py:55

↓ 1 callersFunction_camel2snake

(s)

src/convert_upload/open_clip/hf_model.py:31

↓ 1 callersFunction_compose_two_images

Inserting the second minibatch into the first at the target locations.

src/transforms/mixup.py:293

↓ 1 callersFunction_compute_crop_shape

Compute aspect ratio-preserving shape for central crop. The resulting shape retains `crop_proportion` along one side and a proportion less tha

src/transforms/simclr_aug.py:181

↓ 1 callersFunction_convert_openclip_transformer

(module: Transformer, prefix)

src/convert_upload/open_clip/convert.py:106

↓ 1 callersFunction_create_noun_tokenizer

(vocab_path)

src/transforms/bert_ops.py:74

↓ 1 callersFunction_cutmix

(x, y)

src/transforms/mixup.py:246

↓ 1 callersMethod_cutmix

Applies cutmix.

src/transforms/mixup.py:117

↓ 1 callersMethod_embed_texts

Returns per-class averaged text embeddings.

src/evaluators/proj/image_text/discriminative_classifier.py:343

↓ 1 callersMethod_erase

Erase an area.

src/transforms/random_erasing.py:152

↓ 1 callersFunction_expand_token

(token, batch_size: int)

src/convert_upload/open_clip/transformer.py:315

↓ 1 callersFunction_fill_rectangle

Fills blank area.

src/transforms/random_erasing.py:23

↓ 1 callersMethod_generate_beamsearch

( self, image_inputs, pad_token_id=None, eos_token_id=None,

src/convert_upload/open_clip/coca_model.py:329

↓ 1 callersFunction_get_builder

(dataset, data_dir)

src/datasets/tfds.py:74

↓ 1 callersFunction_get_dataset_info

(builder)

src/evaluators/proj/image_text/retrieval.py:71

↓ 1 callersFunction_get_field_ref

(config_dict, field_name)

src/configs/common.py:136

↓ 1 callersFunction_get_func_args_and_kwargs

(call)

src/helpers/registry.py:115

↓ 1 callersFunction_get_func_name

(expr)

src/helpers/registry.py:105

↓ 1 callersFunction_get_gaussian_kernel_2d

Compute 2D Gaussian kernel given 1D kernels.

src/transforms/three_aug.py:71

↓ 1 callersFunction_get_pad_data

(data)

src/datasets/input_pipeline.py:170

↓ 1 callersMethod_merge_heads

(self, hidden_states, embed_dim)

src/models/text_transformer.py:314

↓ 1 callersMethod_merge_heads

(self, hidden_states, embed_dim)

src/models/common.py:473

↓ 1 callersMethod_mixup

Applies mixup.

src/transforms/mixup.py:160

↓ 1 callersFunction_natural_key

(string_)

src/convert_upload/open_clip/factory.py:29

↓ 1 callersFunction_pad

Explicitly pad a 4-D image. Equivalent to the implicit padding method offered in `tf.nn.conv2d` and `tf.nn.depthwise_conv2d`, but supports no

src/transforms/three_aug.py:22

↓ 1 callersFunction_random_box

Sample a random box of shape [cut_h, cut_w].

src/transforms/mixup.py:267

↓ 1 callersFunction_setup_size

(size, error_msg)

src/convert_upload/open_clip/transform.py:75

↓ 1 callersFunction_split_and_batch

Splits dataset, calls `get_ds` and returns padded + batched datasets.

src/evaluators/proj/image_text/retrieval.py:131

↓ 1 callersFunction_split_and_batch

Splits dataset, calls `get_ds` and returns padded + batched datasets.

src/evaluators/proj/image_text/discriminative_classifier.py:128

↓ 1 callersMethod_tpu_flash_attention

TPU Flash Attention.

src/models/text_transformer.py:317

↓ 1 callersMethod_tpu_flash_attention

Executes the TPU Flash Attention kernel via shard_map.

src/models/text_decoder_v2.py:171

↓ 1 callersMethod_tpu_flash_attention

TPU Flash Attention.

src/models/common.py:476

↓ 1 callersMethod_tpu_flash_attention

TPU Flash Attention.

src/models/text_decoder.py:182

↓ 1 callersMethod_update_labels

(self, images: tf.Tensor, labels: tf.Tensor, lam: tf.Tensor)

src/transforms/mixup.py:178

↓ 1 callersMethodattention

( self, q_x: torch.Tensor, k_x: Optional[torch.Tensor] = None,

src/convert_upload/open_clip/transformer.py:239

↓ 1 callersFunctionautotype

Auto-converts string to bool/int/float if possible.

src/configs/common.py:117

↓ 1 callersFunctionbatched

Batch data into lists of length *n*. The last batch may be shorter. NOTE based on more-itertools impl, to be replaced by python 3.12 itertools.bat

src/convert_upload/open_clip/zero_shot_classifier.py:9

← previous next →101–200 of 716, ranked by callers