Configs.py

import argparse
from argparse import Namespace

class ARGS:
    task_name='classification'
    patch_size=16
    stride=8
    gpt_layers=3
    d_ff=256
    dropout=0.1
    seed=321
    droprate_data=0.2
    model_dir_name='gpt4tsV2'
    model_name='gpt_allzi'
    seq_len=512
    horizon_len=48
    p2n_rate_train=-1
    p2n_rate_test=1
    data_stride_len_train=100
    data_stride_len_test=100
    batch_size=128
    llm_model='GPT2'
    tslm_model='GPT2'
    n_runs=5
    timestamp_type='fixed'
    if_test_save=False
    out_proj='First'
    if_addtoken_grad=False
    if_only_metavec=False
    ifAdapter=False

    # important
    ifP_tuning=False
    ifLoRA=False

    adam_epsilon=1e-08
    learning_rate=1e-05
    # lr_scheduler
    ifschedule_with_warmup=True
    schedule_type='LambdaLR'
    # vec->str 
    ifReprogram=False
    Reprogram_type='ATT'
    # if_decompose
    if_decompose=True
    decompose_type='Simple'
    # ifAdapter
    ifAdapter=False

    if_dont_use_id=False
    ifNoNorm=False
    if_use_his01=False
    if_use_meta=False
    patience=15
    MAX_epoches=200
    npy_dir_path='<DATASET_ROOT>/seqlen_512_horizon_48_stride_50_noPeriod'
    dir_path='<PROJECT_ROOT>'
    kids_path='<DATASET_ROOT>/kids.npy'
    meta_feats_path='<DATASET_ROOT>/kepler_meta_feats3.npy'
    num_classes=2
    dim_in=2
    dim_meta_1=35
    dim_meta_0=3000
    dim_hid=128
    d_model=768


    @classmethod
    def _get_class_variables(cls):
        return {k: v for k, v in vars(cls).items() if not callable(v) and not k.startswith("_")}
    @classmethod
    def _update_class_variables(cls, updates):
        if updates is None:
            return
        if isinstance(updates,Namespace):
            updates=updates.__dict__
        for k, v in updates.items():
            setattr(cls, k, v)