Functions — github.com/ChristofHenkel/kaggle-asl-fingerspelling-1st-place-solution

hub / github.com/ChristofHenkel/kaggle-asl-fingerspelling-1st-place-solution / functions

Functions328 in github.com/ChristofHenkel/kaggle-asl-fingerspelling-1st-place-solution

⨍ Functions 328◇ Types & classes 82 ↳ Endpoints 5

↓ 42 callersFunctionshape_list

Deal with dynamic shape in tensorflow cleanly.

models/mdl_2_tf.py:15

↓ 21 callersMethod__init__

(self, transforms, always_apply=False, p=1.0,)

configs/augmentations.py:1129

↓ 14 callersMethod__init__

(self, dim: int)

models/mdl_1_pt.py:70

↓ 13 callersMethod__init__

(self, dim: int)

models/mdl_2_pt.py:70

↓ 10 callersMethod__init__

(self, axis=-1, name="glu_activation", **kwargs)

models/mdl_2_tf.py:374

↓ 9 callersMethod__init__

(self, config: Speech2TextConfig, **kwargs)

scripts/modeling_tf_speech_to_text2_cache.py:119

↓ 7 callersMethod_shape

(self, tensor: tf.Tensor, seq_len: int, bsz: int)

scripts/modeling_tf_speech_to_text2_cache.py:261

↓ 4 callersMethod__init__

(self)

scripts/convert_cfg_2_to_tf_lite.py:813

↓ 4 callersFunction_expand_mask

Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`.

scripts/modeling_tf_speech_to_text2_cache.py:100

↓ 4 callersFunctionmake_scale

(encoder_dim)

models/mdl_2_pt.py:362

↓ 4 callersFunctionmake_scale

(encoder_dim)

models/mdl_1_pt.py:483

↓ 4 callersFunctionport_weights_to_tf_model

(pt_model, tf_model)

scripts/convert_cfg_2_to_tf_lite.py:45

↓ 3 callersMethod_shape

(self, tensor: tf.Tensor, seq_len: int, bsz: int)

models/mdl_2_tf.py:650

↓ 3 callersFunctionshift_tokens_right

(input_ids: tf.Tensor, pad_token_id: int, decoder_start_token_id: int)

scripts/modeling_tf_speech_to_text2_cache.py:57

↓ 2 callersMethod_get_embedding

Build sinusoidal embeddings. This matches the implementation in tensor2tensor, but differs slightly from the description in Section 3

scripts/modeling_tf_speech_to_text2_cache.py:170

↓ 2 callersMethod_get_feature_vector_attention_mask

(self, feature_vector_length, attention_mask)

scripts/modeling_tf_speech_to_text2_cache.py:750

↓ 2 callersFunctionbatch_to_device

(batch, device)

data/ds_1.py:11

↓ 2 callersMethodcall_attention

(self, query, key, value, logits, training=False, mask=None)

models/mdl_2_tf.py:187

↓ 2 callersMethodcall_qkv

(self, query, key, value, training=False)

models/mdl_2_tf.py:165

↓ 2 callersMethodextend_pe

(self, x)

models/mdl_2_pt.py:132

↓ 2 callersMethodextend_pe

(self, x)

models/mdl_1_pt.py:132

↓ 2 callersMethodgenerate

(self, x, max_new_tokens=33, encoder_attention_mask=None)

models/mdl_2_pt.py:43

↓ 2 callersMethodgenerate

(self, x, max_new_tokens=33, encoder_attention_mask=None)

models/mdl_1_pt.py:41

↓ 2 callersMethodget_config

(self)

models/mdl_2_tf.py:386

↓ 2 callersFunctionget_score

(phrase_gt, phrase_preds)

metrics/metric_1.py:8

↓ 2 callersMethodload_one

(self, file_id, sequence_id)

data/ds_1.py:218

↓ 2 callersMethodload_one

(self, file_id, sequence_id)

data/ds_2.py:244

↓ 2 callersFunctionport_weights_to_tf_model2

(pt_model, tf_model, mapper)

scripts/convert_cfg_2_to_tf_lite.py:220

↓ 2 callersFunctionrotate_half

Rotates half the hidden dims of the input.

models/mdl_2_pt.py:731

↓ 2 callersFunctionspatial_random_affine

(data,scale=None,shear=None,shift=None,degree=None,center=(0,0))

configs/augmentations.py:618

↓ 2 callersFunctiontf_rotate_half

(x)

models/mdl_2_tf.py:614

↓ 1 callersMethod_get_feat_extract_output_lengths

Computes the output length of the convolutional layers

scripts/modeling_tf_speech_to_text2_cache.py:741

↓ 1 callersFunction_make_causal_mask

Make causal mask used for bi-directional self-attention.

scripts/modeling_tf_speech_to_text2_cache.py:82

↓ 1 callersMethod_relative_shift

(self, pos_score: Tensor)

models/mdl_1_pt.py:246

↓ 1 callersFunctionapply_rotary_pos_emb

(q, k, cos, sin)

models/mdl_2_pt.py:737

↓ 1 callersMethodaugment

(self,x)

data/ds_1.py:187

↓ 1 callersMethodaugment

(self,x)

data/ds_2.py:213

↓ 1 callersMethodbuild

(self, input_shape)

models/mdl_2_tf.py:123

↓ 1 callersFunctioncalc_metric

(cfg, pp_out, val_df, pre="val")

metrics/metric_1.py:15

↓ 1 callersMethodcall

(self, x, decoder_input_ids=None, training=False, return_dict=False,**kwargs)

scripts/convert_cfg_2_to_tf_lite.py:689

↓ 1 callersMethodcall

(self, x, decoder_input_ids=None, training=False, return_dict=False,**kwargs)

scripts/convert_cfg_2_to_tf_lite.py:748

↓ 1 callersFunctioncount_parameters

(model)

models/mdl_2_pt.py:57

↓ 1 callersFunctioncount_parameters

(model)

models/mdl_1_pt.py:55

↓ 1 callersMethodcreate_position_ids_from_input_ids

Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols are ignored. This i

scripts/modeling_tf_speech_to_text2_cache.py:213

↓ 1 callersFunctiondecode

(generated_ids)

scripts/get_train_folded_oof_supp.py:42

↓ 1 callersMethodfill_nans

(self,x)

data/ds_1.py:28

↓ 1 callersMethodfill_nans

(self,x)

data/ds_2.py:28

↓ 1 callersMethodfill_nans

(self,x)

scripts/convert_cfg_2_to_tf_lite.py:145

↓ 1 callersFunctionflip

(data, flip_array)

data/ds_1.py:46

↓ 1 callersFunctionflip

(data, flip_array)

data/ds_2.py:44

↓ 1 callersFunctionget_lm_type

(lm)

models/mdl_2_pt.py:748

↓ 1 callersFunctionget_score

(phrase_gt, phrase_preds)

scripts/get_train_folded_oof_supp.py:30

↓ 1 callersFunctioninterpolate_or_pad

(data, max_len=100, mode="start")

data/ds_1.py:51

↓ 1 callersFunctioninterpolate_or_pad

(data, max_len=100, mode="start")

data/ds_2.py:52

↓ 1 callersMethodinterpolate_tf

(self, data, max_len)

scripts/convert_cfg_2_to_tf_lite.py:873

↓ 1 callersMethodnormalize

(self,x)

data/ds_1.py:22

↓ 1 callersMethodnormalize

(self,x)

data/ds_2.py:22

↓ 1 callersMethodnormalize

(self,x)

scripts/convert_cfg_2_to_tf_lite.py:134

↓ 1 callersMethodnormalize_part

(x_in_part_)

scripts/convert_cfg_2_to_tf_lite.py:882

↓ 1 callersFunctionouter_cutmix

(data, phrase, data2, phrase2)

data/ds_1.py:67

↓ 1 callersFunctionouter_cutmix

(data, phrase,score, data2, phrase2,score2)

data/ds_2.py:67

↓ 1 callersFunctionpost_process_pipeline

(cfg, val_data, val_df)

postprocess/pp_2.py:7

↓ 1 callersMethodrelative_shift

(x)

models/mdl_2_tf.py:312

↓ 1 callersFunctionsave_tflite_model

(tf_file, tflite_model_path)

scripts/convert_cfg_2_to_tf_lite.py:1098

↓ 1 callersFunctionset_seed

(seed=1234)

utils.py:22

↓ 1 callersFunctiontf_apply_rotary_pos_emb

(q, k, cos, sin)

models/mdl_2_tf.py:619

↓ 1 callersMethodtokenize

(self,phrase)

data/ds_1.py:191

↓ 1 callersMethodtokenize

(self,phrase)

data/ds_2.py:218

Method__call__

(self, *args, force_apply: bool = False, **data)

configs/augmentations.py:1138

( self, sample_rate=(0.8,1.2), always_apply=False, p=0.5, )

configs/augmentations.py:42

Method__init__

( self, length=384, always_apply=False, p=0.5, )

configs/augmentations.py:88

Method__init__

( self, size=(0.2,0.4), mask_value=float('nan'), num_masks = (1,2), a

configs/augmentations.py:130

Method__init__

( self, size=(0.2,0.4), num_masks = (1,2), always_apply=False, p=0.5,

configs/augmentations.py:183

Method__init__

( self, size=(0.2,0.4), num_masks = (1,2), always_apply=False, p=0.5,

configs/augmentations.py:239

Method__init__

( self, size=(0.2,0.4), mask_value=float('nan'), num_masks = (1,2), a

configs/augmentations.py:294

Method__init__

( self, size=(0.5,1.), mask_value=float('nan'), mode = 'abolute', alw

configs/augmentations.py:361

Method__init__

( self, size=(0.5,1.), mask_value=float('nan'), mode = 'abolute', alw

configs/augmentations.py:432

Method__init__

( self, size=(0.05,0.1), mask_value=0, num_masks = (1,2), mode = 'rel

configs/augmentations.py:503

Method__init__

( self, noise_range=(-0.05,0.05), always_apply=False, p=0.5, )

configs/augmentations.py:585

Method__init__

( self, scale = None, shear = None, shift = None, degree = None,

configs/augmentations.py:678

Method__init__

( self, scale = None, shear = None, shift = None, degree = None,

configs/augmentations.py:746

Method__init__

( self, landmarks, n_fingers, mask_value=float('nan'), always_apply=

configs/augmentations.py:839

Method__init__

(self, transform, landmark_indices, always_apply=False, p=1.0

configs/augmentations.py:893

Method__init__

( self, landmarks, mask_value=float('nan'), always_apply=False, p=1.0

configs/augmentations.py:937

Method__init__

( self, shift_rate=(-10,10), always_apply=False, p=0.5, )

configs/augmentations.py:983

Method__init__

( self, landmarks, mask_value=float('nan'), always_apply=False, p=1.0

configs/augmentations.py:1035

Method__init__

(self, transform, window_size = (0.4,0.6), num_windows = (1,2),

configs/augmentations.py:1076

Method__init__

( self, sample_rate=(0.8,1.2), windows = (5,10), always_apply=False, p

configs/augmentations.py:1171

Method__init__

( self, landmarks, mask_value=float('nan'), always_apply=False, p=1.0

configs/augmentations.py:1229

Method__init__

( self, landmarks, mask_value=float('nan'), always_apply=False, p=1.0

configs/augmentations.py:1286

(self, df, cfg, aug=None, mode="train")

(self, df, cfg, aug=None, mode="train")

data/ds_2.py:92

Method__init__

(self)

scripts/convert_cfg_2_to_tf_lite.py:131

Method__init__

(self, out_dim, n_landmarks)

scripts/convert_cfg_2_to_tf_lite.py:165

Method__init__

( self, decoder_config, name="decoder", **kwargs, )

scripts/convert_cfg_2_to_tf_lite.py:671

Method__init__

( self, decoder_config, name="decoder", **kwargs, )

scripts/convert_cfg_2_to_tf_lite.py:730

next →1–100 of 328, ranked by callers