import os from src.train import train from src.utils import get_experiments_dir, get_data_dir opts = { # ---folders--- "data_path": os.path.join(get_data_dir(), 'SCM', 'free-field'), "logs_path": os.path.join(get_experiments_dir(), 'fc'), "experiment_name": 'fc_freefield_full', # ---network structure--- "model_name": 'fc', "input_sh_order": 3, "rank": None, # None -> output is full matrix, Int -> output is low rank matrix transformed into full matrix "hidden_layers": 1, "hidden_sizes": [3500], "residual_flag": False, "residual_only": False, # ---data--- "batch_size": 25, "num_workers": 15, # ---optimization--- "lr": 1e-3, "lr_sched_thresh": 0.01, "lr_sched_patience": 5, "max_epochs": 1000, "gpus": -1 } train(opts)
from abc import abstractmethod import numpy as np import torch import torch.nn as nn import torch.nn.functional as F from src.data.sig2scm_dataset import Dataset from src.models.base_model import BaseModel from src.utils import get_data_dir, get_experiments_dir import src.utils.complex_torch as ctorch default_opts = { # ---folders--- "data_path": os.path.join(get_data_dir(), 'whitenoise_10_reflections'), "logs_path": get_experiments_dir(), "experiment_name": 'rnn_QA', # ---data options--- "center_frequency": 2500., "bandwidth": 400, # ---network structure--- "model_name": 'rnn', "loss": 'mse', # 'mse' "sh_order_sig": float("inf"), "sh_order_scm": float("inf"), "time_len_sig": float("inf"), # ---data--- # "dtype": torch.float32, # TODO: implement (does errors in saving hyperparameters) "transform": None, "batch_size": 25, "num_workers": 0,
import os from src.train import train from src.utils import get_experiments_dir, get_data_dir opts = { # ---folders--- "data_path": os.path.join(get_data_dir(), 'SCM', 'image-method'), "logs_path": os.path.join(get_experiments_dir(), 'fc', 'image_method'), "experiment_name": 'fc_imagemethod_full', # ---network structure--- "model_name": 'fc', "input_sh_order": 3, "rank": None, # None -> output is full matrix, Int -> output is low rank matrix transformed into full matrix "hidden_layers": 1, "hidden_sizes": [3000], "residual_flag": True, "residual_only": False, # ---data--- "batch_size": 50, "num_workers": 15, # ---optimization--- "lr": 1e-3, "lr_sched_thresh": 0.01, "lr_sched_patience": 10, "max_epochs": 1000, "gpus": -1 } train(opts)