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

↓ 118 callersFunction_pcfg

(url='', hf_hub='', **kwargs)

src/convert_upload/open_clip/pretrained.py:29

↓ 47 callersFunction_n2p

(w, t=True)

src/convert_upload/open_clip/convert.py:21

↓ 28 callersFunctionwrite_note

(note)

src/main_clip.py:128

↓ 16 callersMethodload

(self, ckpt={})

src/helpers/utils.py:406

↓ 15 callersFunctionwrite_note

(note)

src/main_openvision2.py:115

↓ 13 callersFunction_slpcfg

(url='', hf_hub='', **kwargs)

src/convert_upload/open_clip/pretrained.py:42

↓ 13 callersMethodtokenize

(self, text)

src/convert_upload/open_clip/tokenizer.py:534

↓ 11 callersFunctionget_steps

(name, default=ValueError, cfg=config)

src/main_clip.py:211

↓ 11 callersFunctionunbox_logicallypartioned

Unboxes the flax.LogicallyPartitioned pieces Args: boxed_pytree: a pytree that includes LogicallyPartitioned leaves. Returns: a pyt

src/helpers/sharding.py:25

↓ 10 callersMethod__init__

( self, dim: int, num_heads: int = 8, qkv_bias: bool = True,

src/convert_upload/open_clip/transformer.py:90

↓ 10 callersMethodmeasure

Logs the metric value.

src/helpers/utils.py:859

↓ 8 callersFunction_create_bert_tokenizer

(vocab_path, add_bos=False, add_eos=False)

src/transforms/bert_ops.py:44

↓ 8 callersFunctionreshard

Take an arbitrarily* sharded pytree and shard it according to `shardings`. This is a no-op for tree elements which are already sharded as request

src/helpers/sharding.py:59

↓ 8 callersMethodsave

(self)

src/helpers/utils.py:398

↓ 7 callersFunction_apcfg

(url='', hf_hub='', **kwargs)

src/convert_upload/open_clip/pretrained.py:55

↓ 7 callersFunctionitstime

Returns True if it's time to execute an action. Args: step: the current step representing "now". every_n_steps: the action should run every

src/helpers/utils.py:209

↓ 7 callersFunctionrandom_apply

Randomly apply function func to x with probability p.

src/transforms/simclr_aug.py:24

↓ 7 callersMethodrun

Returns metrics.

src/evaluators/proj/image_text/retrieval.py:305

↓ 7 callersFunctiontree_flatten_with_names

Populates tree_flatten with leaf names. This function populates output of tree_flatten with leaf names, using a custom traversal that produces na

src/helpers/utils.py:509

↓ 6 callersFunctioncreate_orbax_checkpoint_manager

Returns specified Orbax (async or not) CheckpointManager or None if checkpointing is disabled.

src/helpers/utils.py:1142

↓ 6 callersFunctionpp

(title, names, indent=" ")

src/models/common.py:590

↓ 6 callersFunctionredux

(x)

src/losses/common.py:86

↓ 6 callersFunctionscale_channel

Scale the 2D image using the autocontrast rule.

src/transforms/autoaugment.py:305

↓ 6 callersFunctiontokenizer_eval

(inkey)

src/configs/openvision2.py:260

↓ 5 callersFunction_make_mask_trees

(params, patterns_values, log)

src/optim/build_optax.py:281

↓ 5 callersFunction_ntuple

(n)

src/convert_upload/open_clip/utils.py:49

↓ 5 callersFunctiondownload_pretrained_from_hf

( model_id: str, filename: str = 'open_clip_pytorch_model.bin', revision=None,

src/convert_upload/open_clip/pretrained.py:616

↓ 5 callersMethodencode_text

(self, text, normalize: bool = False)

src/convert_upload/open_clip/model.py:271

↓ 5 callersFunctionget_model_config

(model_name)

src/convert_upload/open_clip/factory.py:80

↓ 5 callersFunctionmerge_preprocess_dict

Merge overlay key-value pairs on top of base preprocess cfg or dict. Input dicts are filtered based on PreprocessCfg fields.

src/convert_upload/open_clip/transform.py:40

↓ 5 callersMethodpause

(self, wait_for=())

src/helpers/utils.py:389

↓ 5 callersFunctionprefetch_iterator

Runs iterator `it` ahead for `n` steps. Adapted from flax.

src/datasets/input_pipeline.py:192

↓ 5 callersMethodresume

(self)

src/helpers/utils.py:394

↓ 5 callersMethodstep_end

Ends a training step, write its full row.

src/helpers/utils.py:877

↓ 5 callersMethodstep_start

(self, step)

src/helpers/utils.py:855

↓ 5 callersMethodtick

A chronometer tick.

src/helpers/utils.py:320

↓ 5 callersFunctionunwrap

Unwraps an image produced by wrap. Where there is a 0 in the last channel for every spatial position, the rest of the three channels in that spati

src/transforms/autoaugment.py:419

↓ 5 callersFunctionwrap

Returns 'image' with an extra channel set to all 1s.

src/transforms/autoaugment.py:411

↓ 5 callersMethodwrite

(metrics)

src/helpers/utils.py:881

↓ 4 callersFunction_build_vision_tower

( embed_dim: int, vision_cfg: CLIPVisionCfg, quick_gelu: bool = False, cast_dt

src/convert_upload/open_clip/model.py:106

↓ 4 callersFunction_clean_tag

(tag: str)

src/convert_upload/open_clip/pretrained.py:514

↓ 4 callersMethod_encode_image

(self, images, normalize: bool = True)

src/convert_upload/open_clip/coca_model.py:149

↓ 4 callersFunction_get_hf_config

(model_id, cache_dir=None)

src/convert_upload/open_clip/factory.py:87

↓ 4 callersMethod_loss

(self, image_features, text_features, logit_scale, logit_bias=None, negative_only=False)

src/convert_upload/open_clip/loss.py:349

↓ 4 callersMethod_make_layer

(self, planes, blocks, stride=1)

src/convert_upload/open_clip/modified_resnet.py:132

↓ 4 callersFunction_mccfg

(url='', hf_hub='', **kwargs)

src/convert_upload/open_clip/pretrained.py:68

↓ 4 callersFunction_randomly_negate_tensor

With 50% prob turn the tensor negative.

src/transforms/autoaugment.py:473

↓ 4 callersFunction_translate_level_to_arg

(level, translate_const)

src/transforms/autoaugment.py:516

↓ 4 callersFunctionbasic_clean

(text)

src/convert_upload/open_clip/tokenizer.py:67

↓ 4 callersFunctionblend

Blend image1 and image2 using 'factor'. Factor can be above 0.0. A value of 0.0 means only image1 is used. A value of 1.0 means only image2 is us

src/transforms/autoaugment.py:99

↓ 4 callersFunctionblockwise_attn

( query, key, value, bias=None, deterministic=True, dropout_rng=None,

src/models/bpt.py:50

↓ 4 callersMethodclose

(self)

src/helpers/utils.py:889

↓ 4 callersMethoddecode

(self, tokens)

src/convert_upload/open_clip/tokenizer.py:222

↓ 4 callersMethodglobal_registry

()

src/helpers/registry.py:137

↓ 4 callersFunctionhms

Format time in hours/minutes/seconds.

src/helpers/utils.py:254

↓ 4 callersFunctioninfo

(s, *a)

src/main_clip.py:125

↓ 4 callersMethodinform

Provide some extra info that's only known later in the program.

src/helpers/utils.py:304

↓ 4 callersMethodlog_timing

Use this when you time sth once per step and want instant flushing.

src/helpers/utils.py:413

↓ 4 callersMethodsave_pretrained

(self, dest)

src/convert_upload/open_clip/tokenizer.py:427

↓ 4 callersFunctionstartstop_prof

Runs the profiler for `surround` steps around the next `log_steps`.

src/helpers/utils.py:818

↓ 4 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/optim/build_optax.py:24

↓ 3 callersFunction_build_text_tower

( embed_dim: int, text_cfg: CLIPTextCfg, quick_gelu: bool = False, cast_dtype:

src/convert_upload/open_clip/model.py:175

↓ 3 callersMethod_global_pool

(self, x: torch.Tensor)

src/convert_upload/open_clip/transformer.py:600

↓ 3 callersFunction_parse_policy_info

Return the function that corresponds to `name` and update `level` param.

src/transforms/autoaugment.py:578

↓ 3 callersMethod_split_heads

(self, hidden_states, num_heads, head_dim)

src/models/text_transformer.py:311

↓ 3 callersMethod_split_heads

(self, hidden_states, num_heads, head_dim)

src/models/common.py:470

↓ 3 callersFunction_square875

(sz)

src/configs/clip_common.py:47

↓ 3 callersFunctioncreate_mesh

Creates a device mesh with each slice in its own data parallel group.

src/helpers/sharding.py:41

↓ 3 callersMethodencode

(self, text)

src/convert_upload/open_clip/tokenizer.py:214

↓ 3 callersMethodencode_image

(self, image, normalize: bool = False)

src/convert_upload/open_clip/model.py:267

↓ 3 callersFunctionevaluators

()

src/main_clip.py:631

↓ 3 callersFunctionfreeze_batch_norm_2d

Converts all `BatchNorm2d` and `SyncBatchNorm` layers of provided module into `FrozenBatchNorm2d`. If `module` is itself an instance of eithe

src/convert_upload/open_clip/utils.py:9

↓ 3 callersFunctionget_cast_dtype

(precision: str)

src/convert_upload/open_clip/model.py:88

↓ 3 callersFunctionget_clean_fn

(type: str)

src/convert_upload/open_clip/tokenizer.py:94

↓ 3 callersMethodget_logits

(self, image_features, text_features, logit_scale)

src/convert_upload/open_clip/loss.py:102

↓ 3 callersFunctionget_order

(x)

src/transforms/bert_ops.py:59

↓ 3 callersFunctionget_pretrained_cfg

(model: str, tag: str)

src/convert_upload/open_clip/pretrained.py:550

↓ 3 callersMethodget_tfdata

Creates this data object as a tf.data.Dataset. This will be called separately in each process, and it is up to the dataset implementation to

src/datasets/core.py:25

↓ 3 callersFunctionimage_transform_v2

( cfg: PreprocessCfg, is_train: bool, aug_cfg: Optional[Union[Dict[str, Any], Augmenta

src/convert_upload/open_clip/transform.py:393

↓ 3 callersFunctioninfo

(s, *a)

src/main_openvision2.py:112

↓ 3 callersFunctionload_checkpoint

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

src/helpers/utils.py:144

↓ 3 callersMethodlock

lock modules Args: unlocked_groups (int): leave last n layer groups unlocked (default: 0)

src/convert_upload/open_clip/timm_model.py:110

↓ 3 callersFunctionmaybe_repeat

(arg, n_reps)

src/helpers/registry.py:35

↓ 3 callersFunctionmerge_preprocess_kwargs

(base: PreprocessCfg, **kwargs)

src/convert_upload/open_clip/transform.py:57

↓ 3 callersFunctionmin_tensor

(a, b)

src/transforms/autoaugment.py:546

↓ 3 callersFunctionrecover_tree

Recovers a tree as a nested dict from flat names and values. This function is useful to analyze checkpoints that are saved by our programs withou

src/helpers/utils.py:722

↓ 3 callersFunctionshould_merge

(name)

src/models/common.py:565

↓ 3 callersFunctionstartstop_prof_at_steps

( sess, step=None, first_step=None, last_step=None, name="steps", ttl=3 * 365 * 24 * 3600)

src/helpers/utils.py:827

↓ 3 callersFunctiontext_global_pool

(x, text: Optional[torch.Tensor] = None, pool_type: str = 'argmax')

src/convert_upload/open_clip/transformer.py:655

↓ 2 callersMethod__init__

( self, local_loss=False, gather_with_grad=False, cache_labels

src/convert_upload/open_clip/loss.py:68

↓ 2 callersMethod__init__

(self, inplanes, planes, stride=1)

src/convert_upload/open_clip/modified_resnet.py:13

↓ 2 callersMethod__init__

(self, use_pooler_output=True)

src/convert_upload/open_clip/hf_model.py:67

↓ 2 callersFunction_add_tpu_host_options

(data)

src/datasets/input_pipeline.py:185

↓ 2 callersFunction_canonicalize

Returns canonicalized `text` (lowercase and puncuation removed). Args: text: string to be canonicalized. keep_punctuation_exact_string: If

src/evaluators/proj/image_text/prompt_engineering.py:105

↓ 2 callersFunction_convert_openclip_txt

(module: TextTransformer, prefix)

src/convert_upload/open_clip/convert.py:125

↓ 2 callersFunction_convert_timm_img

(module, prefix)

src/convert_upload/open_clip/convert.py:37

↓ 2 callersMethod_embed

Embeds features name `name` using `embed_fn`. Args: name: Feature name to be embedded. train_state: train_state for the predict

src/evaluators/proj/image_text/retrieval.py:223

↓ 2 callersMethod_encode_text

(self, text, normalize: bool = True)

src/convert_upload/open_clip/coca_model.py:154

↓ 2 callersFunction_get_gaussian_kernel

Compute 1D Gaussian kernel.

src/transforms/three_aug.py:62

↓ 2 callersFunction_get_hf_config_tokenizer

(model_id, cache_dir=None)

src/convert_upload/transfer_jax2hf.py:564

next →1–100 of 716, ranked by callers