import os import torch import numpy as np import random import logging import time from absl import flags from torch_geometric.loader import DataLoader import pickle from scipy.stats import pearsonr, spearmanr import wandb import pandas as pd import torch from torch.utils.data import DataLoader #, Subset from models import pgsn from models import cate from models import dagformer from models import digcn from models import digcn_meta from models import regressor from models.GDSS import scorenetx import losses import sampling from models import utils as mutils from models.ema import ExponentialMovingAverage import datasets_nas import sde_lib from utils import * from logger import Logger from analysis.arch_metrics import SamplingArchMetrics, SamplingArchMetricsMeta FLAGS = flags.FLAGS def set_exp_name(config, classifier_config_nf=None): exp_name = f'./exp/{config.task}/{config.folder_name}' wandb_exp_name = exp_name os.makedirs(exp_name, exist_ok=True) config.exp_name = exp_name set_random_seed(config) return exp_name, wandb_exp_name def set_random_seed(config): seed = config.seed os.environ['PYTHONHASHSEED'] = str(seed) torch.manual_seed(seed) torch.cuda.manual_seed(seed) torch.cuda.manual_seed_all(seed) np.random.seed(seed) random.seed(seed) torch.backends.cudnn.deterministic = True torch.backends.cudnn.benchmark = False def sde_train(config): """Runs the training pipeline. Args: config: Configuration to use. workdir: Working directory for checkpoints and TF summaries. If this contains checkpoint training will be resumed from the latest checkpoint. """ # Wandb logger exp_name, wandb_exp_name = set_exp_name(config) wandb_logger = Logger( log_dir=exp_name, exp_name=wandb_exp_name, write_textfile=True, use_wandb=config.log.use_wandb, wandb_project_name=config.log.wandb_project_name) wandb_logger.update_config(config, is_args=True) wandb_logger.write_str(str(vars(config))) wandb_logger.write_str('-' * 100) # Create directories for experimental logs sample_dir = os.path.join(exp_name, "samples") os.makedirs(sample_dir, exist_ok=True) # Initialize model. score_model = mutils.create_model(config) ema = ExponentialMovingAverage(score_model.parameters(), decay=config.model.ema_rate) optimizer = losses.get_optimizer(config, score_model.parameters()) state = dict(optimizer=optimizer, model=score_model, ema=ema, step=0, config=config) # Create checkpoints directly checkpoint_dir = os.path.join(exp_name, "checkpoints") # Intermediate checkpoints to resume training checkpoint_meta_dir = os.path.join(exp_name, "checkpoints-meta", "checkpoint.pth") os.makedirs(checkpoint_dir, exist_ok=True) os.makedirs(os.path.dirname(checkpoint_meta_dir), exist_ok=True) # Resume training when intermediate checkpoints are detected if config.resume: state = restore_checkpoint(config.resume_ckpt_path, state, config.device, resume=config.resume) initial_step = int(state['step']) train_ds, eval_ds, test_ds = datasets_nas.get_dataset(config) train_loader, eval_loader, test_loader = datasets_nas.get_dataloader(config, train_ds, eval_ds, test_ds) n_node_pmf = None # temp print(f'==> # of training elem: {len(train_ds)}') train_iter = iter(train_loader) # create data normalizer and its inverse scaler = datasets_nas.get_data_scaler(config) inverse_scaler = datasets_nas.get_data_inverse_scaler(config) # Setup SDEs if config.training.sde.lower() == 'vpsde': sde = sde_lib.VPSDE(beta_min=config.model.beta_min, beta_max=config.model.beta_max, N=config.model.num_scales) sampling_eps = 1e-3 elif config.training.sde.lower() == 'vesde': sde = sde_lib.VESDE(sigma_min=config.model.sigma_min, sigma_max=config.model.sigma_max, N=config.model.num_scales) sampling_eps = 1e-5 else: raise NotImplementedError(f"SDE {config.training.sde} unknown.") # Build one-step training and evaluation functions optimize_fn = losses.optimization_manager(config) continuous = config.training.continuous reduce_mean = config.training.reduce_mean likelihood_weighting = config.training.likelihood_weighting train_step_fn = losses.get_step_fn(sde, train=True, optimize_fn=optimize_fn, reduce_mean=reduce_mean, continuous=continuous, likelihood_weighting=likelihood_weighting, data=config.data.name) eval_step_fn = losses.get_step_fn(sde, train=False, optimize_fn=optimize_fn, reduce_mean=reduce_mean, continuous=continuous, likelihood_weighting=likelihood_weighting, data=config.data.name) # Build sampling functions if config.training.snapshot_sampling: sampling_shape = (config.training.eval_batch_size, config.data.max_node, config.data.n_vocab) sampling_fn = sampling.get_sampling_fn( config, sde, sampling_shape, inverse_scaler, sampling_eps, config.data.name) num_train_steps = config.training.n_iters # Build analysis tools sampling_metrics = SamplingArchMetrics(config, train_ds, exp_name) # visualization_tools = ArchVisualization(config, remove_none=False, exp_name=exp_name) # -------- Train --------- # logging.info("Starting training loop at step %d." % (initial_step,)) element = {'train': ['training_loss'], 'eval': ['eval_loss'], 'test': ['test_loss'], 'sample': ['r_valid', 'r_unique', 'r_novel'], 'valid_error': ['multi_node_type', 'INVALID_1OR2', 'INVALID_3AND4', 'x_elem_sum']} is_best = False min_test_loss = 1e05 for step in range(initial_step, num_train_steps + 1): try: x, adj, extra = next(train_iter) except StopIteration: train_iter = train_loader.__iter__() x, adj, extra = next(train_iter) mask = aug_mask(adj, algo=config.data.aug_mask_algo, data=config.data.name) x, adj, mask = scaler(x.to(config.device)), adj.to(config.device), mask.to(config.device) # mask = cate_mask(adj) # adj, mask = dense_adj(graphs, config.data.max_node, scaler, config.data.dequantization) batch = (x, adj, mask) # Execute one training step loss = train_step_fn(state, batch) wandb_logger.update(key="training_loss", v=loss.item()) if step % config.training.log_freq == 0: logging.info("step: %d, training_loss: %.5e" % (step, loss.item())) # Report the loss on evaluation dataset periodically if step % config.training.eval_freq == 0: for eval_x, eval_adj, eval_extra in eval_loader: eval_mask = aug_mask(eval_adj, algo=config.data.aug_mask_algo, data=config.data.name) eval_x, eval_adj, eval_mask = scaler(eval_x.to(config.device)), eval_adj.to(config.device), eval_mask.to(config.device) eval_batch = (eval_x, eval_adj, eval_mask) eval_loss = eval_step_fn(state, eval_batch) logging.info("step: %d, eval_loss: %.5e" % (step, eval_loss.item())) wandb_logger.update(key="eval_loss", v=eval_loss.item()) for test_x, test_adj, test_extra in test_loader: test_mask = aug_mask(test_adj, algo=config.data.aug_mask_algo, data=config.data.name) test_x, test_adj, test_mask = scaler(test_x.to(config.device)), test_adj.to(config.device), test_mask.to(config.device) test_batch = (test_x, test_adj, test_mask) test_loss = eval_step_fn(state, test_batch) logging.info("step: %d, test_loss: %.5e" % (step, test_loss.item())) wandb_logger.update(key="test_loss", v=test_loss.item()) if wandb_logger.logs['test_loss'].avg < min_test_loss: is_best = True # Save a checkpoint periodically and generate samples if step != 0 and step % config.training.snapshot_freq == 0 or step == num_train_steps: # Save the checkpoint. save_step = step // config.training.snapshot_freq # save_checkpoint(os.path.join(checkpoint_dir, f'checkpoint_{save_step}.pth'), state) save_checkpoint(checkpoint_dir, state, step, save_step, is_best) # Generate and save samples if config.training.snapshot_sampling: ema.store(score_model.parameters()) ema.copy_to(score_model.parameters()) sample, sample_steps, _ = sampling_fn(score_model, mask) # sample: [batch_size, num_node, n_vocab] sample_list = quantize(sample, adj, alpha=config.sampling.alpha, qtype=config.sampling.qtype) # quantization this_sample_dir = os.path.join(sample_dir, "iter_{}".format(step)) os.makedirs(this_sample_dir, exist_ok=True) # check samples arch_metric = sampling_metrics(arch_list=sample_list, adj=adj, mask=mask, this_sample_dir=this_sample_dir, test=False) r_valid, r_unique, r_novel = arch_metric[0][0], arch_metric[0][1], arch_metric[0][2] if len(arch_metric[0]) > 3: error_type_1 = arch_metric[0][3] error_type_2 = arch_metric[0][4] error_type_3 = arch_metric[0][5] x_elem_sum = int(torch.sum(torch.tensor(sample_list))) else: error_type_1 = None logging.info("step: %d, r_valid: %.5e" % (step, r_valid)) logging.info("step: %d, r_unique: %.5e" % (step, r_unique)) logging.info("step: %d, r_novel: %.5e" % (step, r_novel)) if error_type_1 is not None: logging.info("step: %d, multi_node_type: %.5e" % (step, error_type_1)) logging.info("step: %d, INVALID_1OR2: %.5e" % (step, error_type_2)) logging.info("step: %d, INVALID_3AND4: %.5e" % (step, error_type_3)) logging.info("step: %d, x_elem_sum: %d" % (step, x_elem_sum)) # writer.add_scalar("r_valid", r_valid, step) # res = nasbench201.get_prop(sample_valid_str_list=sample_valid_str) if config.log.use_wandb: # wandb_logger.log_sample(sample) wandb_logger.update(key="r_valid", v=r_valid) wandb_logger.update(key="r_unique", v=r_unique) wandb_logger.update(key="r_novel", v=r_novel) if error_type_1 is not None: wandb_logger.update(key="multi_node_type", v=error_type_1) wandb_logger.update(key="INVALID_1OR2", v=error_type_2) wandb_logger.update(key="INVALID_3AND4", v=error_type_3) wandb_logger.update(key="x_elem_sum", v=x_elem_sum) if config.log.log_valid_sample_prop: wandb_logger.log_valid_sample_prop(arch_metric, x_axis='latency', y_axis='test_acc') if step % config.training.eval_freq == 0: wandb_logger.write_log(element=element, step=step) else: wandb_logger.write_log(element={'train': ['training_loss']}, step=step) wandb_logger.reset() wandb_logger.save_log() def meta_predictor_train(config): # Wandb logger exp_name, wandb_exp_name = set_exp_name(config) wandb_logger = Logger( log_dir=exp_name, exp_name=wandb_exp_name, write_textfile=True, use_wandb=config.log.use_wandb, wandb_project_name=config.log.wandb_project_name) wandb_logger.update_config(config, is_args=True) wandb_logger.write_str(str(vars(config))) wandb_logger.write_str('-' * 100) # Create directories for experimental logs sample_dir = os.path.join(exp_name, "samples") os.makedirs(sample_dir, exist_ok=True) # Initialize model. predictor_model = mutils.create_model(config) optimizer = losses.get_optimizer(config, predictor_model.parameters()) state = dict(optimizer=optimizer, model=predictor_model, step=0, config=config) # Create checkpoints directly checkpoint_dir = os.path.join(exp_name, "checkpoints") # Intermediate checkpoints to resume training checkpoint_meta_dir = os.path.join(exp_name, "checkpoints-meta", "checkpoint.pth") os.makedirs(checkpoint_dir, exist_ok=True) os.makedirs(os.path.dirname(checkpoint_meta_dir), exist_ok=True) # Resume training when intermediate checkpoints are detected state = restore_checkpoint(checkpoint_meta_dir, state, config.device, resume=config.resume) initial_step = int(state['step']) # Build dataloader and iterators train_ds, eval_ds, test_ds = datasets_nas.get_meta_dataset(config) train_loader, eval_loader, test_loader = datasets_nas.get_dataloader(config, train_ds, eval_ds, test_ds) train_iter = iter(train_loader) # create data normalizer and its inverse scaler = datasets_nas.get_data_scaler(config) inverse_scaler = datasets_nas.get_data_inverse_scaler(config) # Setup SDEs if config.training.sde.lower() == 'vpsde': sde = sde_lib.VPSDE(beta_min=config.model.beta_min, beta_max=config.model.beta_max, N=config.model.num_scales) sampling_eps = 1e-3 elif config.training.sde.lower() == 'vesde': sde = sde_lib.VESDE(sigma_min=config.model.sigma_min, sigma_max=config.model.sigma_max, N=config.model.num_scales) sampling_eps = 1e-5 else: raise NotImplementedError(f"SDE {config.training.sde} unknown.") # Build one-step training and evaluation functions optimize_fn = losses.optimization_manager(config) continuous = config.training.continuous reduce_mean = config.training.reduce_mean likelihood_weighting = config.training.likelihood_weighting train_step_fn = losses.get_step_fn_predictor(sde, train=True, optimize_fn=optimize_fn, reduce_mean=reduce_mean, continuous=continuous, likelihood_weighting=likelihood_weighting, data=config.data.name, label_list=config.data.label_list, noised=config.training.noised, t_spot=config.training.t_spot, is_meta=True) eval_step_fn = losses.get_step_fn_predictor(sde, train=False, optimize_fn=optimize_fn, reduce_mean=reduce_mean, continuous=continuous, likelihood_weighting=likelihood_weighting, data=config.data.name, label_list=config.data.label_list, noised=config.training.noised, t_spot=config.training.t_spot, is_meta=True) # Build sampling functions and Load pre-trained score network if config.training.snapshot_sampling: sampling_shape = (config.training.eval_batch_size, config.data.max_node, config.data.n_vocab) sampling_fn = sampling.get_sampling_fn(config, sde, sampling_shape, inverse_scaler, sampling_eps, config.data.name, conditional=True, is_meta=True, data_name='cifar10', num_sample=config.model.num_sample) # Score model score_config = torch.load(config.scorenet_ckpt_path)['config'] check_config(score_config, config) score_model = mutils.create_model(score_config) score_ema = ExponentialMovingAverage(score_model.parameters(), decay=score_config.model.ema_rate) score_state = dict(model=score_model, ema=score_ema, step=0, config=score_config) score_state = restore_checkpoint(config.scorenet_ckpt_path, score_state, device=config.device, resume=True) score_ema.copy_to(score_model.parameters()) num_train_steps = config.training.n_iters # Build analysis tools sampling_metrics = SamplingArchMetricsMeta(config, train_ds, exp_name) # -------- Train --------- # logging.info("Starting training loop at step %d." % (initial_step,)) element = {'train': ['training_loss'], 'eval': ['eval_loss']} is_best = False max_eval_p_corr = -1 for step in range(initial_step, num_train_steps + 1): try: x, adj, extra, task = next(train_iter) except StopIteration: train_iter = train_loader.__iter__() x, adj, extra, task = next(train_iter) mask = aug_mask(adj, algo=config.data.aug_mask_algo, data=config.data.name) x, adj, mask = scaler(x.to(config.device)), adj.to(config.device), mask.to(config.device) # task = task.to(config.device) if config.data.name == 'NASBench201' else [_.to(config.device) for _ in task] task = [_.to(config.device) for _ in task] if config.data.name == 'ofa' else task.to(config.device) # mask = cate_mask(adj) # adj, mask = dense_adj(graphs, config.data.max_node, scaler, config.data.dequantization) batch = (x, adj, mask, extra, task) # Execute one training step loss, pred, labels = train_step_fn(state, batch) wandb_logger.update(key="training_loss", v=loss.item()) if step % config.training.log_freq == 0: logging.info("step: %d, training_loss: %.5e" % (step, loss.item())) # Save a temporary checkpoint to resume training after pre-emption periodically if step != 0 and step % config.training.snapshot_freq_for_preemption == 0: save_checkpoint(checkpoint_meta_dir, state, step, save_step, is_best) # Report the loss on evaluation dataset periodically if step % config.training.eval_freq == 0: eval_pred_list, eval_labels_list = list(), list() for eval_x, eval_adj, eval_extra, eval_task in eval_loader: eval_mask = aug_mask(eval_adj, algo=config.data.aug_mask_algo, data=config.data.name) eval_x, eval_adj, eval_mask = scaler(eval_x.to(config.device)), eval_adj.to(config.device), eval_mask.to(config.device) eval_task = [_.to(config.device) for _ in eval_task] eval_batch = (eval_x, eval_adj, eval_mask, eval_extra, eval_task) eval_loss, eval_pred, eval_labels = eval_step_fn(state, eval_batch) eval_pred_list += [v.detach().item() for v in eval_pred.squeeze()] eval_labels_list += [v.detach().item() for v in eval_labels.squeeze()] logging.info("step: %d, eval_loss: %.5e" % (step, eval_loss.item())) wandb_logger.update(key="eval_loss", v=eval_loss.item()) eval_p_corr = pearsonr(np.array(eval_pred_list), np.array(eval_labels_list))[0] eval_s_corr = spearmanr(np.array(eval_pred_list), np.array(eval_labels_list))[0] if eval_p_corr > max_eval_p_corr: is_best = True max_eval_p_corr = eval_p_corr # Save a checkpoint periodically and generate samples if step != 0 and step % config.training.snapshot_freq == 0 or step == num_train_steps: # Save the checkpoint. save_step = step // config.training.snapshot_freq save_checkpoint(checkpoint_dir, state, step, save_step, is_best) # Generate and save samples if config.training.snapshot_sampling: score_ema.store(score_model.parameters()) score_ema.copy_to(score_model.parameters()) sample, sample_steps, sample_chain, (score_grad_norm_p, classifier_grad_norm_p, score_grad_norm_c, classifier_grad_norm_c) = \ sampling_fn(score_model, mask, predictor_model, eval_chain=False, number_chain_steps=config.sampling.number_chain_steps, classifier_scale=config.sampling.classifier_scale) sample_list = quantize(sample, adj) # quantization this_sample_dir = os.path.join(sample_dir, "iter_{}".format(step)) os.makedirs(this_sample_dir, exist_ok=True) arch_metric = sampling_metrics(arch_list=sample_list, adj=adj, mask=mask, this_sample_dir=this_sample_dir, test=False, check_dataname=config.sampling.check_dataname) r_valid, r_unique, r_novel = arch_metric[0][0], arch_metric[0][1], arch_metric[0][2] test_acc_list = arch_metric[2]['test_acc_list'] if step % config.training.eval_freq == 0: wandb_logger.write_log(element=element, step=step) else: wandb_logger.write_log(element={'train': ['training_loss']}, step=step) wandb_logger.reset() def check_config(config1, config2): assert config1.training.sde == config2.training.sde assert config1.training.continuous == config2.training.continuous assert config1.data.centered == config2.data.centered assert config1.data.max_node == config2.data.max_node assert config1.data.n_vocab == config2.data.n_vocab run_train_dict = { 'sde': sde_train, 'meta_predictor': meta_predictor_train } def train(config): run_train_dict[config.model_type](config)