run.py

import os
import sys
import warnings
warnings.filterwarnings("ignore")
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'

import numpy as np
import torch
import tensorflow as tf

# Add parent directory to path
sys.path.append(os.path.join(os.path.dirname('__file__'), '../'))

# Local imports
from experiment import run
from utils.context_fid import Context_FID
from utils.cross_correlation import CrossCorrelLoss
from utils.metric_utils import display_scores
from utils.discriminative_metric import discriminative_score_metrics
from utils.predictive_metric import predictive_score_metrics

# Suppress warnings

# Configure GPU memory growth
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
    try:
        for gpu in gpus:
            tf.config.experimental.set_memory_growth(gpu, True)
    except RuntimeError as e:
        print(e)

# Global settings
iterations = 5
enable_fid = 0
enable_corr = 0
enable_dis = 1
enable_pred = 1

# all_results = {}
# for config_path in [
#     "./config/modified/sines.yaml",
#     "./config/modified/revenue-baseline-365.yaml",
#     "./config/modified/energy.yaml",
#     "./config/modified/fmri.yaml",

#     "./config/modified/96/energy.yaml",
#     "./config/modified/192/energy.yaml",
#     "./config/modified/384/energy.yaml",

#     "./config/modified/96/fmri.yaml",
#     "./config/modified/192/fmri.yaml",
#     "./config/modified/384/fmri.yaml",
    
#     "./config/modified/96/sines.yaml",
#     "./config/modified/192/sines.yaml",
#     "./config/modified/384/sines.yaml",
    
#     "./config/modified/192/revenue.yaml",
#     "./config/modified/96/revenue.yaml",
#     "./config/modified/384/revenue.yaml",
# ]:
#     class Args:
#         config_path = config_path
#         gpu = 0

#     results, dataset_name, seq_length = run(Args())
#     all_results[config_path] = (results, dataset_name, seq_length)

# python run.py ./config/modified/energy.yaml
# python run.py ./config/modified/fmri.yaml
# python run.py ./config/modified/sines.yaml
# python run.py ./config/modified/revenue-baseline-365.yaml

# python run.py ./config/modified/96/energy.yaml
# python run.py ./config/modified/192/energy.yaml
# python run.py ./config/modified/384/energy.yaml

# python run.py ./config/modified/96/fmri.yaml
# python run.py ./config/modified/192/fmri.yaml
# python run.py ./config/modified/384/fmri.yaml

# python run.py ./config/modified/96/sines.yaml
# python run.py ./config/modified/192/sines.yaml
# python run.py ./config/modified/384/sines.yaml

# python run.py ./config/modified/192/revenue.yaml
# python run.py ./config/modified/96/revenue.yaml
# python run.py ./config/modified/384/revenue.yaml

ds_name_display = {
    "sines": "Sine",
    "revenue": "Revenue",
    "energy": "ETTh",
    "fmri": "fMRI",
}

def random_choice(size, num_select=100):
    select_idx = np.random.randint(low=0, high=size, size=(num_select,))
    return select_idx

def compute_metrics(ori_data, fake_data, iterations=5, data_name='sines', data_len=24, key="unconditional"):
    
    if enable_dis:
        discriminative_score = []
        for i in range(iterations):
            temp_disc, fake_acc, real_acc = discriminative_score_metrics(ori_data[:], fake_data[:ori_data.shape[0]])
            discriminative_score.append(temp_disc)
            print(f'Iter {i}: ', temp_disc, ',', fake_acc, ',', real_acc, '\n')
            
        mean, sigma = display_scores(discriminative_score)
        content = f'disc {data_name} {key} {data_len} {mean} {sigma}'

        with open(f'log {data_name}.txt', 'a+') as file:
            file.write(content + '\n')

    if enable_pred:
        predictive_score = []
        for i in range(iterations):
            temp_pred = predictive_score_metrics(ori_data, fake_data[:ori_data.shape[0]])
            predictive_score.append(temp_pred)
            print(i, ' epoch: ', temp_pred, '\n')

        mean, sigma = display_scores(predictive_score)
        content = f'pred {data_name} {key} {data_len} {mean} {sigma}'

        with open(f'log {data_name}.txt', 'a+') as file:
            file.write(content + '\n')

    if enable_fid:
        context_fid_score = []

        for i in range(iterations):
            context_fid = Context_FID(ori_data[:], fake_data[:ori_data.shape[0]])
            context_fid_score.append(context_fid)
            print(f'Iter {i}: ', 'context-fid =', context_fid, '\n')

        mean, sigma = display_scores(context_fid_score)
        content = f'fid {data_name} {key} {data_len} {mean} {sigma}'

        with open(f'log {data_name}.txt', 'a+') as file:
            file.write(content + '\n')


    if enable_corr:
        x_real = torch.from_numpy(ori_data)
        x_fake = torch.from_numpy(fake_data)
        correlational_score = []
        size = int(x_real.shape[0] / iterations)

        for i in range(iterations):
            real_idx = random_choice(x_real.shape[0], size)
            fake_idx = random_choice(x_fake.shape[0], size)
            corr = CrossCorrelLoss(x_real[real_idx, :, :], name='CrossCorrelLoss')
            loss = corr.compute(x_fake[fake_idx, :, :])
            correlational_score.append(loss.item())
            print(f'Iter {i}: ', 'cross-correlation =', loss.item(), '\n')

        mean, sigma = display_scores(correlational_score)
        content = f'corr {data_name} {key} {data_len} {mean} {sigma}'

        with open(f'log {data_name}.txt', 'a+') as file:
            file.write(content + '\n')



import argparse
parser = argparse.ArgumentParser()
parser.add_argument('config_path', type=str, default="./config/modified/sines.yaml")
parser.add_argument('--gpu', type=int, default=0)
#     class Args:
#         config_path = config_path
#         gpu = 0
results, dataset_name, seq_length = run(parser.parse_args())
# config_path = parser.parse_args().config_path
# results, dataset_name, seq_length = all_results[config_path] 

ori_data = results["ori_data"]
unconditional_data = results["unconditional"]
sum_controled_data = results["sum_controlled"]
anchor_controled_data = results["anchor_controlled"]

compute_metrics(ori_data, unconditional_data, iterations=iterations, data_name=ds_name_display[dataset_name], data_len=seq_length, key="unconditional")
for key, value in sum_controled_data.items():
    compute_metrics(ori_data, value, iterations=iterations, data_name=ds_name_display[dataset_name], data_len=seq_length, key=key)
for key, value in anchor_controled_data.items():
    compute_metrics(ori_data, value, iterations=iterations, data_name=ds_name_display[dataset_name], data_len=seq_length, key=key)