Functions — github.com/amazon-science/AnoLLM-large-language-models-for-tabular-anomaly-detection

CodeHub MCPcopy Create free account

hub / github.com/amazon-science/AnoLLM-large-language-models-for-tabular-anomaly-detection / functions

Functions88 in github.com/amazon-science/AnoLLM-large-language-models-for-tabular-anomaly-detection

⨍ Functions 88 ◇ Types & classes 13

↓ 5 callersMethoddecision_function

Obtain anomaly scores for each sample in the test data df_test: pandas dataframe of test data n_permutations: number of permutations to calculate

anollm/anollm.py:206

↓ 5 callersMethodfit

Fine-tune AnoLLM using tabular data. Args: data: Pandas DataFrame that contains the tabular data column_names: If data is Numpy Array, the fe

anollm/anollm.py:115

↓ 5 callersFunctionload_data

src/data_utils.py:550

↓ 4 callersMethodget_column_names

anollm/anollm_dataset.py:45

↓ 3 callersMethod__init__

(self, hidden_sizes, num_bins = 7)

src/baselines/dte.py:18

↓ 3 callersFunctionconvert_np_to_df

src/data_utils.py:542

↓ 3 callersFunctiondf_to_numpy

( X: pd.DataFrame, dataset_name: Optional[str] = None, method: Optional[str] = 'ordinal', normalize_

src/data_utils.py:665

↓ 3 callersMethodfit

(self, X_train, y_train = None, X_test = None, y_test = None, verbose=False)

src/baselines/dte.py:106

↓ 3 callersMethodfix_column_order

anollm/anollm_dataset.py:55

↓ 3 callersMethodforward_noise

Takes data point and a timestep as input and returns the noisy version of it

src/baselines/dte.py:83

↓ 3 callersFunctionget_metrics

(args, only_raw = False, only_normalized = False, only_ordinal = False)

src/get_results.py:78

↓ 3 callersFunctionget_run_name

evaluate_baselines.py:51

↓ 3 callersFunctionget_text_columns

src/data_utils.py:641

↓ 3 callersFunctionis_baseline

src/get_results.py:137

↓ 3 callersMethodprepare

Preprocess the data by tokenizing each column and truncating the columns to max_length Inputs: max_length_dict specifies the maximum length of

anollm/anollm_dataset.py:60

↓ 3 callersMethodset_textual_columns

(self, columns: tp.List[str])

anollm/anollm_dataset.py:34

↓ 3 callersMethodset_tokenizer

Set the Tokenizer Args: tokenizer: Tokenizer from HuggingFace

anollm/anollm_dataset.py:22

↓ 2 callersFunction_array_to_dataframe

Converts a Numpy Array to a Pandas DataFrame Args: data: Pandas DataFrame or Numpy NDArray columns: If data is a Numpy Array, col

anollm/anollm_utils.py:5

↓ 2 callersMethod_getitem

Get one instance of the tabular data, permuted, converted to text and tokenized.

anollm/anollm_dataset.py:125

↓ 2 callersFunctionaggregate_results

src/get_results.py:161

↓ 2 callersFunctionbenchmark

evaluate_baselines.py:66

↓ 2 callersFunctionget_max_length_dict

src/data_utils.py:651

↓ 2 callersFunctionget_model

train_anollm.py:119

↓ 2 callersMethodget_n_columns

anollm/anollm_dataset.py:41

↓ 2 callersFunctionget_run_name

train_anollm.py:66

↓ 2 callersFunctionload_adbench_data

src/data_utils.py:478

↓ 2 callersFunctionload_dataset

(dataset_name, data_dir)

src/data_utils.py:82

↓ 2 callersFunctionscores_calc_internal

(query, positive,no_negatives,tau)

src/baselines/icl.py:17

↓ 2 callersMethodshuffle_column_order

anollm/anollm_dataset.py:49

↓ 2 callersFunctionsplit_data

( X: pd.DataFrame, dataset_name: str, n_splits: int, data_dir: str, train_ratio: Optional[float]

src/data_utils.py:493

↓ 1 callersMethod__init__

(self, seed=0, model_name="ICL", num_epochs = 2000, no_batchs = 3000, no_negatives=1000, temperature=0.1, lr=0

src/baselines/icl.py:131

↓ 1 callersFunctiona_minus_b

src/baselines/icl.py:73

↓ 1 callersFunctionbinning

Gives the bin number for a given t based on T (maximum) and the number of bins This is floor(t*num_bins/T) bounded by 0 and T-1

src/baselines/dte.py:49

↓ 1 callersMethodcompute_loss

src/baselines/dte.py:103

↓ 1 callersFunctiondata_generator

(subsample=None, target_label=None)

src/data_utils.py:400

↓ 1 callersMethoddecision_function

src/baselines/dte.py:141

↓ 1 callersFunctionfilter_results

src/get_results.py:142

↓ 1 callersFunctionflatten

src/data_utils.py:409

↓ 1 callersFunctionget_args

evaluate_anollm.py:16

↓ 1 callersFunctionget_args

evaluate_baselines.py:34

↓ 1 callersFunctionget_args

train_anollm.py:16

↓ 1 callersFunctionget_args

src/get_results.py:11

↓ 1 callersFunctionget_args

src/get_avg_results.py:9

↓ 1 callersMethodload_from_state_dict

Load AnoLLM model from state_dict Args: path: path where AnoLLM model is saved

anollm/anollm.py:316

↓ 1 callersFunctionmain

evaluate_anollm.py:53

↓ 1 callersFunctionmain

evaluate_baselines.py:174

↓ 1 callersFunctionmain

train_anollm.py:92

↓ 1 callersFunctionmain

src/get_results.py:201

↓ 1 callersFunctionmain

src/get_avg_results.py:24

↓ 1 callersFunctionnormalize

(X, method, n_buckets)

src/data_utils.py:582

↓ 1 callersFunctionpositive_matrice_builder

(dataset, kernel_size)

src/baselines/icl.py:43

↓ 1 callersFunctionprint_dataset_information

(dataset, data_dir)

src/data_utils.py:795

↓ 1 callersMethodsave_state_dict

Save AnoLLM Model Saves the model weights and a configuration file in the given directory. Warning: Only works in DDP setting! Args: path:

anollm/anollm.py:297

↓ 1 callersMethodset_anomaly_label

anollm/anollm_dataset.py:30

↓ 1 callersMethodset_eval_setting

(self, n_permutations)

anollm/anollm_trainer.py:51

↓ 1 callersFunctionsplit_on_uppercase

src/data_utils.py:241

↓ 1 callersFunctiontabular_metrics

Calculates evaluation metrics for tabular anomaly detection. Adapted from https://github.com/xuhongzuo/DeepOD/blob/main/deepod/metrics/_anom

src/get_results.py:27

↓ 1 callersFunctiontake_per_row_complement

(A, indx, num_elem=3)

src/baselines/icl.py:33

FunctionIP_to_country

src/data_utils.py:334

(self, features: tp.List[tp.Dict[str, tp.Any]])

anollm/anollm_dataset.py:207

Method__getitem__

src/baselines/icl.py:87

Method__getitems__

(self, keys: tp.Union[int, slice, str, list])

anollm/anollm_dataset.py:191

Args: llm: HuggingFace checkpoint of a pretrained large language model, used a basis of our model experiment_dir: Directory, where the trai

anollm/anollm.py:46

(self, seed = 0, model_name = "DTE", hidden_size = [256, 512, 256], epochs = 400, batch_size = 64, lr = 1e-4,

src/baselines/dte.py:57

(self, seed = 0, model_name = "DTE_categorical", hidden_size = [256, 512, 256], epochs = 400, batch_size = 64,

src/baselines/dte.py:163

(self, seed = 0, model_name = "DTE_inverse_gamma", hidden_size = [256, 512, 256], epochs = 400, batch_size = 6

src/baselines/dte.py:187

(self, seed = 0, model_name = "DTE_gaussian", hidden_size = [256, 512, 256], epochs = 400, batch_size = 64, lr

src/baselines/dte.py:224

src/baselines/icl.py:81

(self, kernel_size,hdn_size,d)

src/baselines/icl.py:94

src/baselines/icl.py:84

Function_seed_worker

Helper function to set worker seed during Dataloader initialization.

anollm/anollm_trainer.py:151

Methodcompute_loss

src/baselines/dte.py:170

Methodcompute_loss

src/baselines/dte.py:190

Methodcompute_loss

src/baselines/dte.py:227

Methoddecision_function

src/baselines/dte.py:207

Methoddecision_function

src/baselines/icl.py:211

(self, eval_dataset=None, ignore_keys=None, metric_key_prefix: str = "eval")

anollm/anollm_trainer.py:54

Functionf1_calculator

(classes, losses)

src/baselines/icl.py:61

Functionfilter_anomalies

(X_test, y_test)

src/data_utils.py:811

(self,X_train, y_train=None)

src/baselines/icl.py:150

src/baselines/dte.py:40

src/baselines/icl.py:112

Methodget_item_test

anollm/anollm_dataset.py:175

Methodget_train_dataloader

anollm/anollm_trainer.py:27

Functionget_word

src/data_utils.py:597

Functionordinal

src/data_utils.py:586

Methodset_epoch

anollm/anollm_dataset.py:222

Functiontake_per_row

(A, indx, num_elem=2)

src/baselines/icl.py:56