Functions — github.com/athms/learning-from-brains

↓ 1 callersFunctionget_args

()

scripts/analyses/fig2_upstream_data_overview.py:148

↓ 1 callersFunctionget_args

()

scripts/analyses/sfig7_downstream_learning_curves_replication.py:27

↓ 1 callersFunctionget_args

()

scripts/analyses/fig3_hyperopt.py:183

↓ 1 callersFunctionget_args

()

scripts/analyses/sfig6_downstream_performance_replication.py:27

↓ 1 callersFunctionget_args

()

scripts/analyses/sfig2_eval_error_brain_maps.py:205

↓ 1 callersFunctionget_args

()

scripts/dataprep/upstream/dataprep.py:156

↓ 1 callersFunctionget_args

()

scripts/dataprep/downstream/hcp_dataprep.py:268

↓ 1 callersFunctionget_args

()

scripts/dataprep/downstream/mdtb_dataprep.py:273

↓ 1 callersFunctionget_args

()

scripts/dataprep/downstream/hcp_to_bids.py:405

↓ 1 callersFunctionget_pretrained_architectures_with_all_train_styles

()

tests/__init__.py:70

↓ 1 callersFunctionget_pretrained_architectures_with_pretrain_styles

()

tests/__init__.py:64

↓ 1 callersFunctionget_test_functions

()

tests/__init__.py:151

↓ 1 callersFunctiongrab_tarfile_paths

(path)

src/tools/data.py:8

↓ 1 callersFunctionhcp_to_bids

(args: argparse.Namespace=None)

scripts/dataprep/downstream/hcp_to_bids.py:12

↓ 1 callersFunctionidentify_event_type_from_filepath

( filepath: str, task: str )

scripts/dataprep/downstream/hcp_to_bids.py:230

↓ 1 callersMethodis_next_loss

( self, is_next_logits, is_next )

src/embedder/bert.py:191

↓ 1 callersMethodis_next_loss

( self, is_next_logits, is_next )

src/embedder/netbert.py:191

↓ 1 callersMethodis_next_pred

( self, outputs: torch.tensor )

src/decoder/bert.py:39

↓ 1 callersFunctionlaunched_by_deepspeed

()

tests/__init__.py:96

↓ 1 callersMethodloss

( self, batch, outputs )

src/embedder/base.py:265

↓ 1 callersFunctionmake_decoder

Make a decoder object. The decoder contains the core model architecture used for learning. Args: ----- architecture: st

src/decoder/make.py:6

↓ 1 callersFunctionmake_embedder

Make an embedder object. The embedder is used to prepare an input batch (as generated by src.batcher) for training and compute

src/embedder/make.py:6

↓ 1 callersFunctionmake_trainer

Make a Trainer object for training a model. Returns an instance of transformers.Trainer. See the HuggingFace transformers documentat

src/trainer/make.py:99

↓ 1 callersFunctionmake_unembedder

Creates a UnEmbedder object. Args ---- embed_dim: int Dimension of the embedding space. out_dim: int Dimension o

src/unembedder.py:103

↓ 1 callersFunctionmap_tasks_to_labels

Map task names to numeric labels.

scripts/dataprep/downstream/mdtb_dataprep.py:219

↓ 1 callersFunctionplot_brain_map

( img, path, mask_img=None, workdir=None, axsize=(7, 5), dpi=300, w_pad=0, h_p

src/tools/brainmap.py:195

↓ 1 callersFunctionpreprocess_hcp

Script's main function; additional preprocessing of HCP fmriprep derivatives

scripts/dataprep/downstream/hcp_dataprep.py:14

↓ 1 callersFunctionpreprocess_mdtb

Script's main function; additional preprocessing of MDTB fmriprep derivatives

scripts/dataprep/downstream/mdtb_dataprep.py:15

↓ 1 callersFunctionquality_check

test of source BOLD data for fmriprep-20.2.0 bug. see: https://github.com/nipreps/fmriprep/issues/2307

scripts/dataprep/upstream/dataprep.py:145

↓ 1 callersMethodreconstruction_loss

( self, input, target, **kwargs )

src/embedder/dummy.py:48

↓ 1 callersMethodset_confounds_strategy

( self, strategy: List[str]=None, motion: str="basic", wm_csf: str="basic",

src/preprocessor.py:346

↓ 1 callersMethodset_difumo

( self, dimension=1024, resolution=2 )

src/preprocessor.py:271

↓ 1 callersMethodset_masker

( self, smoothing_fwhm: int=3, standardize: str='zscore', detrend: bool=True,

src/preprocessor.py:292

↓ 1 callersMethodset_template

( self, resolution=2 )

src/preprocessor.py:258

↓ 1 callersFunctionsfig_downstream_learning_curves_replication

Script's main funtion; creates Appendix Figure 7 by wrapping fig_downstream_performance() from scripts/analyses/fig5_downstream-performance.py

scripts/analyses/sfig7_downstream_learning_curves_replication.py:11

↓ 1 callersFunctionsfig_downstream_performance_replication

Script's main funtion; creates Appendix Figure 6 by wrapping fig_downstream_performance() from scripts/analyses/fig5_downstream-performance.py

scripts/analyses/sfig6_downstream_performance_replication.py:11

↓ 1 callersFunctionsfig_training_curve_seqBERT_large

Script's main function; creates Appendix Figure 1 of the manuscript.

scripts/analyses/sfig1_training_curve_largest_sequence_BERT.py:19

↓ 1 callersFunctionsfig_upstream_performance_pretrained_lms

Script's main function; creates Appendix Figure 3 of the manuscript.

scripts/analyses/sfig3_upstream_performance_pretrained_lms.py:19

↓ 1 callersMethodstack_inputs

( self, tensor )

src/unembedder.py:66

↓ 1 callersFunctionsummarize_evs

( ev_filepaths: List[str], task: str, subject: Union[int, str], run: Union[int, str] )

scripts/dataprep/downstream/hcp_to_bids.py:267

↓ 1 callersMethodswitch_decoding_mode

Switch model to decoding model or back to training mode. Necessary to adapt pre-trained models to downstream decoding tasks.

src/model.py:85

↓ 1 callersFunctionto_cl_args

(config)

tests/__init__.py:110

↓ 1 callersMethodunstack_inputs

( self, tensor, b )

src/unembedder.py:76

↓ 1 callersFunctionyiel_rest_ev

Yield random intervals (between seq_min and seq_max) for resting state data.

scripts/dataprep/downstream/hcp_dataprep.py:236

↓ 1 callersFunctionyield_task_ev

Yield type, onset, and end of events in task EV file.

scripts/dataprep/downstream/hcp_dataprep.py:220

↓ 1 callersFunctionyield_task_ev

Yield type, onset, and end of events in task EV file.

scripts/dataprep/downstream/mdtb_dataprep.py:175

Method__getitem__

(self, idx)

src/batcher/bert.py:29

Method__getitem__

(self, idx)

src/batcher/base.py:46

Method__init__

(self, root: str, dataset: str, derivatives_path: str=None, verbose: bool=Fals

src/preprocessor.py:19

Method__init__

( self, embed_dim: int = 768, out_dim: int = 1024, num_hidden_layers: int = 1,

src/unembedder.py:31

Method__init__

( self, embedder: torch.nn.Module, decoder: torch.nn.Module, unembedder: torch

src/model.py:35

Method__init__

( self, **kwargs )

src/embedder/csm.py:11

Method__init__

( self, masking_rate: float = 0.2, n_positions: int=512, **kwargs )

src/embedder/mnm.py:11

Method__init__

( self, **kwargs )

src/embedder/autoen.py:7

Method__init__

( self, **kwargs )

src/embedder/bert.py:11

Method__init__

( self, in_dim: int = 1024, embed_dim: int = 768, num_hidden_layers: int = 1,

src/embedder/base.py:10

Method__init__

(self, **kwargs)

src/embedder/dummy.py:8

Method__init__

( self, **kwargs )

src/embedder/netbert.py:11

Method__init__

( self, masking_rate: float = 0.2, **kwargs )

src/embedder/msm.py:11

Method__init__

( self, n_networks: int=1024, n_timesteps: int=50, num_decoding_classes: int=1

src/decoder/linear_baseline.py:9

Method__init__

( self, **kwargs )

src/decoder/bert.py:80

Method__init__

( self, num_hidden_layers: int = 4, embed_dim: int = 768, dropout: float = 0.2

src/decoder/autoencoder.py:111

Method__init__

( self, num_hidden_layers: int = 4, embed_dim: int = 768, output_dim: int = 10

src/decoder/autoencoder.py:260

Method__init__

( self, **kwargs )

src/decoder/gpt.py:149

Method__init__

( self, is_deepspeed: bool = False, **kwargs )

src/trainer/base.py:9

Method__init__

(self)

src/trainer/make.py:16

Method__init__

( self, gap_min: int = 1, gap_max: int = 5, **kwargs )

src/batcher/bert.py:251

Method__init__

( self, dataloader, length, sample_keys )

src/batcher/base.py:32

Method__init__

( self, sample_random_seq: bool = True, seq_min: int = 10, seq_max: int = 50,

src/batcher/base.py:62

Method__len__

(self)

src/batcher/bert.py:26

Method__len__

(self)

src/batcher/base.py:43

Function_cat_data_collator

(features: List)

src/trainer/make.py:73

Method_root_loss

( self, masked_inputs, outputs, modelling_mask, **kwargs )

src/embedder/csm.py:228

Method_root_loss

( self, masked_inputs, is_next, outputs, is_next_logits, model

src/embedder/bert.py:204

Method_root_loss

( self, masked_inputs, is_next, outputs, is_next_logits, model

src/embedder/netbert.py:204

Methodcompute_loss

Compute training loss, based on embedder's training-style. Args ---- batch: Dict[str, torch.tensor]

src/model.py:108

Functionconfigure_wandb

( config, entity: str=None, project: str='learning-from-brains', mode: str='online', set_w

src/tools/wandb.py:7

Methodcount_source_bold

(self)

src/preprocessor.py:191

Methoddataset

Create Pytorch dataset that can be used for training. Args: ----- tarfiles: list List of paths to data fi

src/batcher/base.py:114

Methoddecode

( self, outputs: Dict[str, torch.tensor], )

src/decoder/linear_baseline.py:66

Functiondecoding_accuracy_metrics

(eval_preds)

src/trainer/make.py:90

Methodembed_inputs

( self, inputs: torch.tensor )

src/embedder/dummy.py:22

Methodforward

( self, inputs, **kwargs )

src/unembedder.py:88

Methodforward

( self, inputs, **kwargs )

src/embedder/base.py:69

Methodforward

( self, batch: Dict[str, torch.tensor] )

src/embedder/base.py:187

Methodforward

( self, batch: Dict[str, torch.tensor] )

src/embedder/dummy.py:28

Methodforward

( self, batch: Dict[str, torch.tensor] )

src/decoder/linear_baseline.py:72

Methodforward

( self, batch: Dict[str, torch.tensor] )

src/decoder/bert.py:58

Methodforward

( self, inputs: torch.tensor = None, hidden: torch.tensor = None, cell_state:

src/decoder/autoencoder.py:55

Methodforward

( self, inputs: torch.tensor, targets: torch.tensor, hidden: torch.tensor = No

src/decoder/autoencoder.py:212

Methodforward

( self, batch: Dict[str, torch.tensor], )

src/decoder/autoencoder.py:332

Methodforward

( self, batch: Dict[str, torch.tensor] )

src/decoder/gpt.py:122

Methodloss

( self, batch, outputs )

src/embedder/dummy.py:101

Methodmask_inputs

( self, batch: Dict[str, torch.tensor] )

src/embedder/mnm.py:40

Methodmask_inputs

( self, batch: Dict[str, torch.tensor] )

src/embedder/msm.py:23

Functionmodel_init

(params: Dict=None)

scripts/train.py:220

Methodon_log

( self, args, state, control, model, **kwargs )

src/trainer/make.py:21

Functionplot_model_graph

( model, dataloader, path: str=None )

src/tools/visualize.py:6

Methodprediction_step

( self, model, batch, prediction_loss_only: bool = False, ignore_keys:

src/trainer/base.py:18

Methodprep_batch

( self, batch: Dict[str, torch.tensor], )

src/embedder/csm.py:45

Functions222 in github.com/athms/learning-from-brains