MCPcopy Create free account

hub / github.com/crowsonkb/cloob-training / functions

Functions58 in github.com/crowsonkb/cloob-training

↓ 7 callersMethod__init__
(self, approximate=True)
cloob_training/model_pt.py:21
↓ 5 callersMethod__init__
(self, d_ff, name=None)
cloob_training/model.py:31
↓ 4 callersFunctionhopfield
(state_patterns, stored_patterns, scale_hopfield)
cloob_training/loss.py:38
↓ 4 callersFunctionunreplicate
(x)
train.py:29
↓ 2 callersFunctioninfoloob_loss
Computes the InfoLOOB loss (negative mean log odds assigned to positive pairs).
cloob_training/loss.py:19
↓ 2 callersFunctionmap_to_tensor
(x)
cloob_training/model_pt.py:181
↓ 1 callersFunctioncloob_loss
Computes the CLOOB loss (negative mean log odds assigned to positive pairs after Hopfield retrieval). Note: this loss has been rescaled f
cloob_training/loss.py:5
↓ 1 callersFunctionconvert_jax_text_params
(params)
cloob_training/model_pt.py:233
↓ 1 callersFunctionconvert_jax_vit_image_params
(params)
cloob_training/model_pt.py:185
↓ 1 callersFunctioncosine_decay_schedule
Implements an cosine learning rate schedule with an optional exponential warmup. init_value (float): The initial learning rate. st
train.py:85
↓ 1 callersFunctionget_and_init_model
(config, key)
cloob_training/model.py:129
↓ 1 callersFunctionget_hk_model
(config)
cloob_training/model.py:107
↓ 1 callersFunctionhopfield_retrieval
(image_features, text_features, scale_hopfield)
cloob_training/loss.py:30
↓ 1 callersFunctioninverse_decay_schedule
Implements an inverse decay learning rate schedule with an optional exponential warmup. steps is the number of steps required for the learnin
train.py:58
↓ 1 callersFunctionload_config
(path)
cloob_training/config.py:4
↓ 1 callersFunctionmain
()
train.py:115
↓ 1 callersFunctionmain
()
make_slim_checkpoint.py:9
↓ 1 callersFunctionmake_weight_decay_mask
(params)
train.py:106
↓ 1 callersFunctionpsplit
(x, n)
train.py:25
↓ 1 callersFunctionsave
()
train.py:268
↓ 1 callersFunctionset_precision
(mode)
cloob_training/model.py:142
↓ 1 callersFunctiontrain_one_epoch
(params, opt_state)
train.py:254
Method__call__
(self, texts)
train.py:41
Method__call__
(self, batch)
train.py:53
Method__call__
(self, x)
cloob_training/model.py:12
Method__call__
(self, x, padding_mask=None)
cloob_training/model.py:22
Method__call__
(self, x)
cloob_training/model.py:35
Method__call__
(self, x, padding_mask=None)
cloob_training/model.py:50
Method__call__
(self, x)
cloob_training/model.py:67
Method__call__
(self, x)
cloob_training/model.py:91
Method__call__
(self, x, padding_mask=None)
cloob_training/model_pt.py:81
Method__init__
(self, context_length=77)
train.py:34
Method__init__
(self, init=None, name=None)
cloob_training/model.py:8
Method__init__
(self, num_heads=1, name=None)
cloob_training/model.py:18
Method__init__
(self, d_ff, n_heads, name=None)
cloob_training/model.py:45
Method__init__
(self, d_embed, n_layers, d_model, n_heads, vocab_size, name=None)
cloob_training/model.py:57
Method__init__
(self, d_embed, n_layers, d_model, n_heads, patch_size, name=None)
cloob_training/model.py:82
Method__init__
(self, seq_len, d_model)
cloob_training/model_pt.py:12
Method__init__
(self, d_model, n_heads=1)
cloob_training/model_pt.py:32
Method__init__
(self, d_model, d_ff)
cloob_training/model_pt.py:60
Method__init__
(self, d_model, d_ff, n_heads)
cloob_training/model_pt.py:76
Method__init__
(self, d_embed, n_layers, d_model, seq_len, n_heads, vocab_size)
cloob_training/model_pt.py:88
Method__init__
(self, d_embed, n_layers, d_model, seq_len, n_heads, input_channels, patch_size)
cloob_training/model_pt.py:117
Method__init__
(self, config, image_encoder, text_encoder)
cloob_training/model_pt.py:147
Functiondownload_checkpoint
(config)
cloob_training/pretrained.py:22
Methodforward
(self, x)
cloob_training/model_pt.py:16
Methodforward
(self, x)
cloob_training/model_pt.py:25
Methodforward
(self, x, padding_mask=None)
cloob_training/model_pt.py:42
Methodforward
(self, x)
cloob_training/model_pt.py:67
Methodforward
(self, x)
cloob_training/model_pt.py:103
Methodforward
(self, x)
cloob_training/model_pt.py:133
Functionget_config
(name)
cloob_training/pretrained.py:16
Functionget_pt_model
(config)
cloob_training/model_pt.py:158
Functionget_pt_params
(config, checkpoint)
cloob_training/model_pt.py:279
Functionlist_configs
()
cloob_training/pretrained.py:12
Functionloss_fn
(params, inputs, axis_name='i')
train.py:238
Functionschedule
(count)
train.py:79
Functiontrain_step
(params, opt_state, inputs, axis_name='i')
train.py:245