diff --git a/bin/blend_mask_s21.py b/bin/blend_mask_s21.py
deleted file mode 100755
index b331e25e..00000000
--- a/bin/blend_mask_s21.py
+++ /dev/null
@@ -1,54 +0,0 @@
-#!/usr/bin/env python3
-
-import os
-
-import cv2
-import numpy as np
-import tqdm
-
-from saicinpainting.evaluation.data import PrecomputedInpaintingResultsDataset
-from saicinpainting.evaluation.utils import load_yaml
-
-
-def main(args):
-    config = load_yaml(args.config)
-
-    if not args.predictdir.endswith('/'):
-        args.predictdir += '/'
-
-    dataset = PrecomputedInpaintingResultsDataset(args.datadir, args.predictdir, **config.dataset_kwargs)
-
-    os.makedirs(os.path.dirname(args.outpath), exist_ok=True)
-
-    for img_i in tqdm.trange(len(dataset)):
-        try:
-            pred_fname = dataset.pred_filenames[img_i]
-            cur_out_fname = os.path.join(args.outpath, pred_fname[len(args.predictdir):])
-            os.makedirs(os.path.dirname(cur_out_fname), exist_ok=True)
-
-            sample = dataset[img_i]
-            img = sample['image']
-            mask = sample['mask']
-            inpainted = sample['inpainted']
-
-            cur_res = (1 - mask) * img + mask * inpainted
-
-            cur_res = np.clip(np.transpose(cur_res, (1, 2, 0)) * 255, 0, 255).astype('uint8')
-            cur_res = cv2.cvtColor(cur_res, cv2.COLOR_RGB2BGR)
-            cv2.imwrite(cur_out_fname, cur_res)
-        except Exception as ex:
-            print(f'Failed to process {img_i}, {pred_fname} due to {ex}')
-
-
-if __name__ == '__main__':
-    import argparse
-
-    aparser = argparse.ArgumentParser()
-    aparser.add_argument('config', type=str, help='Path to evaluation config')
-    aparser.add_argument('datadir', type=str,
-                         help='Path to folder with images and masks (output of gen_mask_dataset.py)')
-    aparser.add_argument('predictdir', type=str,
-                         help='Path to folder with predicts (e.g. predict_hifill_baseline.py)')
-    aparser.add_argument('outpath', type=str, help='Where to put results')
-
-    main(aparser.parse_args())
diff --git a/bin/gen_val_samples_s21.py b/bin/gen_val_samples_s21.py
deleted file mode 100755
index 9ba16181..00000000
--- a/bin/gen_val_samples_s21.py
+++ /dev/null
@@ -1,65 +0,0 @@
-#!/usr/bin/env python3
-
-import logging
-import os
-import sys
-import traceback
-
-from saicinpainting.evaluation.utils import move_to_device
-
-os.environ['OMP_NUM_THREADS'] = '1'
-os.environ['OPENBLAS_NUM_THREADS'] = '1'
-os.environ['MKL_NUM_THREADS'] = '1'
-os.environ['VECLIB_MAXIMUM_THREADS'] = '1'
-os.environ['NUMEXPR_NUM_THREADS'] = '1'
-
-import cv2
-import hydra
-import numpy as np
-import torch
-import tqdm
-import yaml
-from omegaconf import OmegaConf
-from torch.utils.data._utils.collate import default_collate
-
-from saicinpainting.training.data.datasets import make_default_val_dataset
-from saicinpainting.training.trainers import load_checkpoint
-from saicinpainting.utils import register_debug_signal_handlers
-
-LOGGER = logging.getLogger(__name__)
-
-
-@hydra.main(config_path='../configs/prediction', config_name='default.yaml')
-def main(predict_config: OmegaConf):
-    try:
-        register_debug_signal_handlers()  # kill -10 <pid> will result in traceback dumped into log
-
-        if not predict_config.indir.endswith('/'):
-            predict_config.indir += '/'
-
-        dataset = make_default_val_dataset(predict_config.indir, **predict_config.dataset)
-
-        with torch.no_grad():
-            for img_i in tqdm.trange(len(dataset)):
-                mask_fname = dataset.mask_filenames[img_i]
-                cur_out_fname = os.path.join(predict_config.outdir, os.path.splitext(mask_fname[len(predict_config.indir):])[0] + '.png')
-                os.makedirs(os.path.dirname(cur_out_fname), exist_ok=True)
-
-                sample = dataset[img_i]
-                img = sample['image']
-                mask = sample['mask']
-
-                cur_res = (1 - mask) * img + mask * np.array([0, 1.0, 0])[..., None, None]
-
-                cur_res = np.clip(np.transpose(cur_res, (1, 2, 0)) * 255, 0, 255).astype('uint8')
-                cur_res = cv2.cvtColor(cur_res, cv2.COLOR_RGB2BGR)
-                cv2.imwrite(cur_out_fname, cur_res)
-    except KeyboardInterrupt:
-        LOGGER.warning('Interrupted by user')
-    except Exception as ex:
-        LOGGER.critical(f'Prediction failed due to {ex}:\n{traceback.format_exc()}')
-        sys.exit(1)
-
-
-if __name__ == '__main__':
-    main()
diff --git a/bin/parse_logs_and_download.py b/bin/parse_logs_and_download.py
deleted file mode 100644
index dee17b70..00000000
--- a/bin/parse_logs_and_download.py
+++ /dev/null
@@ -1,164 +0,0 @@
-import os
-import re
-import pathlib
-
-import yaml
-import pandas as pd
-from tqdm import tqdm
-
-ROOT_PATH = "/mnt/storage/images"
-CLUSTER_EXP_PATH = "/vol1/dbstore/orc_srr/multimodal/a.mashikhin/inpainting/experiments/"
-folders = [
-    # list of folders
-]
-
-
-def save_txt_metrics(folder, metrics_block, params):
-    metrics_str = "Validation metrics after epoch #"+metrics_block[:-1] + f"\n\n Params: {round(params,1)}M"
-    path = os.path.join(ROOT_PATH, folder, f"best_metrics.txt")
-    text_file = open(path, "w")
-    text_file.write(metrics_str)
-    text_file.close()
-
-
-def convert_log_to_metrics(file):
-    df = []
-    for i, metrics_block in enumerate(file.split("saicinpainting.training.trainers.base][INFO] - Validation metrics after epoch #")):
-        if i == 0:
-            continue
-        ts = re.findall('\[(.*?)\]', metrics_block)[0]
-
-        metrics_block_raw = metrics_block.split(ts)[0]
-        epoch = metrics_block_raw.split(", total")[0]
-        metrics_list = metrics_block_raw.split("\n")[-2].split()
-        d = {"epoch": int(epoch),
-             "fid_mean": float(metrics_list[1]),
-             "fid_std": float(metrics_list[2]),
-             "lpips_mean": float(metrics_list[3]),
-             "lpips_std": float(metrics_list[4]),
-             "ssim_mean": float(metrics_list[5]),
-             "ssim_std": float(metrics_list[6]),
-             "ssim_fid100_f1_mean": float(metrics_list[7]),
-             "ssim_fid100_f1_std": float(metrics_list[8]),
-            }
-        df.append(d)
-    df = pd.DataFrame(df).drop_duplicates("epoch")
-    df = df.sort_values("ssim_fid100_f1_mean", ascending=False)
-    return df, metrics_block_raw
-
-
-def download_latest_logs(folder):
-    output_path = os.path.join(ROOT_PATH, folder)
-    pathlib.Path(output_path).mkdir(exist_ok=True, parents=True)
-    exp_folder = os.path.join(CLUSTER_EXP_PATH, folder)
-    os.system(f"rsync -av --delete korea:{exp_folder}/train.log {output_path}")
-    if not os.path.exists(os.path.join(output_path, "config.yaml")):
-        os.system(f"rsync -av korea:{exp_folder}/config.yaml {output_path}")
-        
-
-def download_best_images(folder, epoch):
-    print(f"Downloading sample for {folder}, epoch = {epoch}")
-    output_path = os.path.join(ROOT_PATH, folder, "samples")
-    pathlib.Path(output_path).mkdir(exist_ok=True, parents=True)
-    exp_folder = os.path.join(CLUSTER_EXP_PATH, folder, "samples", f"epoch{epoch:04d}_test")
-    os.system(f"rsync -av korea:{exp_folder} {output_path}")
-
-
-def main():
-    result = {}
-    for folder in tqdm(folders):
-        exp_name = folder
-        download_latest_logs(folder)
-        path = os.path.join(ROOT_PATH, folder, "train.log")
-        with open(path) as f:
-            file = f.read()
-        
-        # metrics
-        metrics_df, metrics_block = convert_log_to_metrics(file)
-        best_metrics = metrics_df.iloc[0].to_dict()
-        d = best_metrics
-        
-        # model
-        params = file.split("generator      | ")[1].split("discriminator")[0].strip().replace("|", "").split()[1].strip()
-        yaml_config_path = os.path.join(ROOT_PATH, folder, 'config.yaml')
-        with open(yaml_config_path, 'r') as stream:
-            config = yaml.safe_load(stream)
-        params = float(params)
-        d.update({"ngf": config['generator']['ngf'],
-                "n_blocks": config['generator']['n_blocks'],
-                "kind": config['generator']['kind'],
-                "dilation_num" : config['generator'].get("multidilation_kwargs", {}).get("dilation_num", 0),
-                "params": params})
-        result[exp_name] = d
-        save_txt_metrics(folder, metrics_block, params)
-        download_best_images(folder, int(best_metrics['epoch']))
-
-
-def plot()
-    import matplotlib.pyplot as plt
-    import matplotlib
-    import matplotlib.patches as mpatches
-    %matplotlib inline
-    plt.style.use(['default'])
-    matplotlib.rcParams.update({'font.size': 22})
-
-    x = []
-    y = []
-    plot_df = []
-    for k,v in result.items():
-        new_v = {"name": k}
-        new_v.update(v)
-        plot_df.append(new_v)
-    plot_df = pd.DataFrame(plot_df)
-
-
-    def get_label_and_color(kind, n_block, ngf, dilation_num):
-        if  kind == "pix2pixhd_global":
-            k=""
-            color = "green"
-        elif kind == "pix2pixhd_multidilated":
-            k="_dil"
-            if dilation_num == 2:
-                post = "2"
-                color = "red"
-            elif dilation_num == 4:
-                post = "4"
-                color = "purple"
-            else:
-                raise NotImplementedError( f"dilation_num= {dilation_num}")
-            k = f"{k}{post}"
-        else:
-            raise NotImplementedError(kind)
-        label = f"b{n_block}_n{ngf}{k}"
-        return label, color
-
-    tmp_df = plot_df[plot_df['epoch'] > 0].sort_values("params")
-    x = tmp_df['params'].tolist()
-    y = tmp_df['ssim_fid100_f1_mean'].tolist()
-    kind = tmp_df['kind'].tolist()
-    n_block = tmp_df['n_blocks'].tolist()
-    ngf = tmp_df['ngf'].tolist()
-    dilation_num = tmp_df['dilation_num'].tolist()
-    name = tmp_df['name'].tolist()
-
-    f,ax = plt.subplots(1,1,figsize=(20,12))
-
-    ax.scatter(x,y, color="black") 
-    for i in range(len(tmp_df)):
-        label, color = get_label_and_color(kind=kind[i], n_block=n_block[i], ngf=ngf[i], dilation_num=dilation_num[i])
-        p = round(y[i]/0.9043 *100,1)
-        ax.annotate(f"{label}({p}%)", (x[i], y[i]), fontsize=18, color=color)
-
-    ax.set_xlabel("params (M)")
-    ax.set_ylabel("ssim_fid100_f1")
-    ax.set_title("Distillation of SOTA (ssim_fid100_f1 =0.9043)")
-    patches = [mpatches.Patch(color='green', label='Normal convs'),
-            mpatches.Patch(color='red', label='Dilated convs CATin (dil_num=2)'),
-            mpatches.Patch(color='purple', label='Dilated convs CATin (dil_num=4). (all <= 6 epochs)'),]
-    ax.legend(handles=patches, loc=4)
-    ax.grid()
-    plt.savefig("1.png", format='png', bbox_inches='tight')
-
-
-if __name__ == "__main__":
-    main()
\ No newline at end of file
diff --git a/bin/process_debug.sh b/bin/process_debug.sh
deleted file mode 100755
index 31edbeec..00000000
--- a/bin/process_debug.sh
+++ /dev/null
@@ -1,20 +0,0 @@
-#!/bin/bash
-
-DIRNAME="$(dirname $0)"
-
-$DIRNAME/gen_mask_dataset.py $DIRNAME/../configs/gen_dataset1.yaml \
-    /data/inpainting/Places365/debug \
-    /data/inpainting/Places365/debug_masks1 \
-    --n-jobs 6
-
-$DIRNAME/calc_dataset_stats.py /data/inpainting/Places365/debug_masks1 \
-    /data/inpainting/Places365/debug_masks1_summary \
-    --samples-n 20
-
-$DIRNAME/predict_hifill_baseline.py /data/inpainting/Places365/debug_masks1 \
-    /data/inpainting/Places365/debug_masks1_hifill
-
-$DIRNAME/evaluate_predicts.py $DIRNAME/../configs/eval1.yaml \
-    /data/inpainting/Places365/debug_masks1 \
-    /data/inpainting/Places365/debug_masks1_hifill \
-    /data/inpainting/Places365/metrics/debug_masks1_hifill.csv
diff --git a/bin/process_test_large.sh b/bin/process_test_large.sh
deleted file mode 100755
index f7f3f84f..00000000
--- a/bin/process_test_large.sh
+++ /dev/null
@@ -1,19 +0,0 @@
-#!/bin/bash
-
-DIRNAME="$(dirname $0)"
-
-$DIRNAME/gen_mask_dataset.py $DIRNAME/../configs/gen_dataset1.yaml \
-    /data/inpainting/Places365/test_large \
-    /data/inpainting/Places365/test_large_masks1 \
-    --n-jobs 6
-
-$DIRNAME/calc_dataset_stats.py /data/inpainting/Places365/test_large_masks1 \
-    /data/inpainting/Places365/test_large_masks1_summary
-
-$DIRNAME/predict_hifill_baseline.py /data/inpainting/Places365/test_large_masks1 \
-    /data/inpainting/Places365/test_large_masks1_hifill
-
-$DIRNAME/evaluate_predicts.py $DIRNAME/../configs/eval1.yaml \
-    /data/inpainting/Places365/test_large_masks1 \
-    /data/inpainting/Places365/test_large_masks1_hifill \
-    /data/inpainting/Places365/metrics/test_large_masks1_hifill.csv
diff --git a/bin/process_val_large.sh b/bin/process_val_large.sh
deleted file mode 100755
index 564e8fd8..00000000
--- a/bin/process_val_large.sh
+++ /dev/null
@@ -1,20 +0,0 @@
-#!/bin/bash
-
-DIRNAME="$(dirname $0)"
-
-$DIRNAME/gen_mask_dataset.py $DIRNAME/../configs/gen_dataset1.yaml \
-    /data/inpainting/Places365/val_large \
-    /data/inpainting/Places365/val_large_masks1 \
-    --n-jobs 4
-
-$DIRNAME/calc_dataset_stats.py /data/inpainting/Places365/val_large_masks1 \
-    /data/inpainting/Places365/val_large_masks1_summary \
-    --samples-n 20
-
-$DIRNAME/predict_hifill_baseline.py /data/inpainting/Places365/val_large_masks1 \
-    /data/inpainting/Places365/val_large_masks1_hifill
-
-$DIRNAME/evaluate_predicts.py $DIRNAME/../configs/eval1.yaml \
-    /data/inpainting/Places365/val_large_masks1 \
-    /data/inpainting/Places365/val_large_masks1_hifill \
-    /data/inpainting/Places365/metrics/val_large_masks1_hifill.csv
diff --git a/bin/process_val_large_random.sh b/bin/process_val_large_random.sh
deleted file mode 100755
index 91171b78..00000000
--- a/bin/process_val_large_random.sh
+++ /dev/null
@@ -1,19 +0,0 @@
-#!/bin/bash
-
-DIRNAME="$(dirname $0)"
-DATA_ROOT=/data/inpainting/Places365
-
-for config in random_medium_256 random_medium_512 random_thick_256 random_thick_512 random_thin_256 random_thin_512
-do
-    MASKS_OUTPATH=$DATA_ROOT/val_large_random_masks/$config
-
-    $DIRNAME/gen_mask_dataset.py $DIRNAME/../configs/data_gen/$config.yaml \
-        $DATA_ROOT/val_large \
-        $MASKS_OUTPATH \
-        --n-jobs 4
-
-    $DIRNAME/calc_dataset_stats.py $MASKS_OUTPATH \
-        ${MASKS_OUTPATH}_summary \
-        --samples-n 20
-done
-
diff --git a/bin/sinlge_image_segmentation.py b/bin/sinlge_image_segmentation.py
deleted file mode 100644
index dc4fa050..00000000
--- a/bin/sinlge_image_segmentation.py
+++ /dev/null
@@ -1,190 +0,0 @@
-import argparse
-import os
-import pickle
-import time
-import random
-
-import cv2
-import colorsys
-
-import faiss
-
-import numpy as np
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-import torchvision
-import torchvision.transforms as transforms
-import torchvision.datasets as datasets
-
-import clustering
-
-
-IMAGENET_MEAN = torch.FloatTensor([0.485, 0.456, 0.406])[None, :, None, None]
-IMAGENET_STD  = torch.FloatTensor([0.229, 0.224, 0.225])[None, :, None, None]
-
-
-class VGG19(nn.Module):
-    def __init__(self, net='pytorch', normalize_inputs=True):
-        super(VGG19, self).__init__()
-
-        self.normalize_inputs = normalize_inputs
-        self.mean_ = IMAGENET_MEAN
-        self.std_ = IMAGENET_STD
-
-        vgg = torchvision.models.vgg19(pretrained=True).features
-
-        for weights in vgg.parameters():
-            weights.requires_grad = False
-
-        vgg_avg_pooling = []
-        for module in vgg.modules():
-            if module.__class__.__name__ == 'Sequential':
-                continue
-            elif module.__class__.__name__ == 'MaxPool2d':
-                vgg_avg_pooling.append(nn.AvgPool2d(kernel_size=1, stride=1, padding=0))
-            else:
-                vgg_avg_pooling.append(module)
-
-        self.vgg = nn.Sequential(*vgg_avg_pooling[:-2])
-
-    def do_normalize_inputs(self, x):
-        return (x - self.mean_.to(x.device)) / self.std_.to(x.device)
-
-    def forward(self, input):
-        if self.normalize_inputs:
-            input = self.do_normalize_inputs(input)
-        output = self.vgg(input)
-        return output
-
-
-def random_colors(N, bright=True):
-    """
-    Generate random colors.
-    To get visually distinct colors, generate them in HSV space then
-    convert to RGB.
-    """
-    brightness = 1.0 if bright else 0.7
-    hsv = [(float(i) / N, 1, brightness) for i in range(N)]
-    colors = list(map(lambda c: colorsys.hsv_to_rgb(*c), hsv))
-    random.shuffle(colors)
-    return colors
-
-
-def denormalize_inputs(x):
-    return x * IMAGENET_STD.to(x.device) + IMAGENET_MEAN.to(x.device)
-
-def parse_args():
-    parser = argparse.ArgumentParser(description='PyTorch Implementation of DeepCluster')
-
-    parser.add_argument('--data',       type=str, help='path to dataset')
-    parser.add_argument('--clustering', type=str, choices=['Kmeans', 'PIC'], default='Kmeans')
-    parser.add_argument('--num_cluster', '--k', type=int, default=15)
-    parser.add_argument('--sobel', action='store_true', help='Sobel filtering')
-    
-    parser.add_argument('--batch', default=10, type=int)
-    parser.add_argument('--workers', default=4, type=int)
-
-    return parser.parse_args()
-
-
-def main(args):
-    print('-------------- args --------------')
-    print(args)
-
-
-    print('-------------- MODEL --------------')
-    model = VGG19().cuda()
-    print(model)
-
-
-    print('------------- ImageNet -------------')
-    normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
-                                     std=[0.229, 0.224, 0.225])
-    tra = [transforms.Resize(256),
-           transforms.CenterCrop(224),
-           transforms.ToTensor(),
-           normalize]
-
-    end = time.time()
-    dataset = datasets.ImageFolder(args.data, transform=transforms.Compose(tra))
-    print('Load dataset: {0:.2f} s'.format(time.time() - end))
-
-    dataloader = torch.utils.data.DataLoader(dataset,
-                                             batch_size=args.batch,
-                                             num_workers=args.workers,
-                                             pin_memory=True)
-
-
-    print('-------------- feature infer -----------')  
-
-    for i, (input_tensor, _) in enumerate(dataloader):
-        with torch.no_grad():
-            input_var = input_tensor.cuda()
-            features  = model.forward(input_var)
-
-            print(  'inputs:', input_var.shape, torch.min(input_var), torch.max(input_var))
-            print('features:', features.shape)
-
-        break
-
-    b,c,h,w = input_var.size()
-    images  = 255 * denormalize_inputs(input_var).permute(2,0,3,1).contiguous() # H x B x W x C
-    images  = images.view(h, -1, c).cpu().numpy().astype(np.uint8)
-
-    b,c,h,w  = features.size()
-    features = features.view(b, c, -1).transpose(1,2).contiguous() # B x H*W x C
-    
-    xv, yv = torch.meshgrid([torch.arange(0,h), torch.arange(0,w)]) # H x W , H x W
-    xv = xv.view(1, h, w, 1) 
-    yv = yv.view(1, h, w, 1)
-    coords = torch.cat([xv,yv], dim=3).cuda() # 1 x H x W x 2 
-    coords = coords.expand(size=(b,h,w,2)).view(b,-1, 2).contiguous().float() # B x H*W x 2
-
-    coords = coords / (h*w)
-
-    print('  images:', images.shape)
-    print('features:', features.shape)
-    print('  coords:', coords.shape)
-
-    #features = torch.cat([features, coords], dim=-1)
-
-    print('  concat:', features.shape)
-
-
-    print('-------------- Clustering -----------')
-
-    segmentations = []
-    for idx in range(b):
-        feature = features[idx].cpu().numpy().astype(np.float32)
-        #coord   = coords[idx].cpu().numpy()
-        coord   = None
-
-        clusterizer = clustering.__dict__[args.clustering](args.num_cluster)
-        I, loss     = clusterizer.cluster(data=feature, 
-                                          coord_grid=coord, 
-                                          verbose=True)
-        cluster_map = np.array(I).reshape(h,w)
-
-        segmentation = np.zeros(shape=(h,w,3))
-        lbl_colors   = random_colors(args.num_cluster)
-        for j, color in enumerate(lbl_colors):
-            segmentation[cluster_map==j] = color
-        segmentation = (255*segmentation).astype(np.uint8)
-
-        print('segmentation:', segmentation.shape)
-        segmentations.append(segmentation)
-
-    segmentations = np.concatenate(segmentations, axis=1) # along Width
-
-    print('segmentations:', segmentations.shape)
-
-
-    result = np.concatenate([images,segmentations], axis=0)
-      
-    cv2.imwrite(f'batch0_img.png', result)
-    
-
-if __name__ == '__main__':
-    args = parse_args()
-    main(args=args)
diff --git a/bin/train_benchmark.py b/bin/train_benchmark.py
deleted file mode 100755
index 37499832..00000000
--- a/bin/train_benchmark.py
+++ /dev/null
@@ -1,76 +0,0 @@
-#!/usr/bin/env python3
-
-import logging
-import os
-import sys
-import traceback
-
-os.environ['OMP_NUM_THREADS'] = '1'
-os.environ['OPENBLAS_NUM_THREADS'] = '1'
-os.environ['MKL_NUM_THREADS'] = '1'
-os.environ['VECLIB_MAXIMUM_THREADS'] = '1'
-os.environ['NUMEXPR_NUM_THREADS'] = '1'
-
-import hydra
-from omegaconf import OmegaConf
-from pytorch_lightning import Trainer
-from pytorch_lightning.callbacks import ModelCheckpoint
-from pytorch_lightning.loggers import TensorBoardLogger
-from pytorch_lightning.plugins import DDPPlugin
-
-from saicinpainting.training.trainers import make_training_model
-from saicinpainting.utils import register_debug_signal_handlers, \
-    handle_ddp_subprocess, handle_ddp_parent_process, \
-    handle_deterministic_config
-
-
-from torch.profiler import profile, record_function, ProfilerActivity
-
-LOGGER = logging.getLogger(__name__)
-
-
-@handle_ddp_subprocess()
-@hydra.main(config_path='../configs/training', config_name='tiny_test.yaml')
-def main(config: OmegaConf):
-    try:
-        need_set_deterministic = handle_deterministic_config(config)
-
-        register_debug_signal_handlers()  # kill -10 <pid> will result in traceback dumped into log
-
-        is_in_ddp_subprocess = handle_ddp_parent_process()
-
-        config.visualizer.outdir = os.path.join(os.getcwd(), config.visualizer.outdir)
-        if not is_in_ddp_subprocess:
-            LOGGER.info(OmegaConf.to_yaml(config))
-            OmegaConf.save(config, os.path.join(os.getcwd(), 'config.yaml'))
-
-        checkpoints_dir = os.path.join(os.getcwd(), 'models')
-        os.makedirs(checkpoints_dir, exist_ok=True)
-
-        # there is no need to suppress this logger in ddp, because it handles rank on its own
-        metrics_logger = TensorBoardLogger(config.location.tb_dir, name=os.path.basename(os.getcwd()))
-        metrics_logger.log_hyperparams(config)
-
-        training_model = make_training_model(config)
-        trainer_kwargs = OmegaConf.to_container(config.trainer.kwargs, resolve=True)
-        if need_set_deterministic:
-            trainer_kwargs['deterministic'] = True
-
-        trainer = Trainer(
-            # there is no need to suppress checkpointing in ddp, because it handles rank on its own
-            callbacks=ModelCheckpoint(dirpath=checkpoints_dir, **config.trainer.checkpoint_kwargs),
-            logger=metrics_logger,
-            default_root_dir=os.getcwd(),
-            **trainer_kwargs
-        )
-        trainer.fit(training_model)
-    
-    except KeyboardInterrupt:
-        LOGGER.warning('Interrupted by user')
-    except Exception as ex:
-        LOGGER.critical(f'Training failed due to {ex}:\n{traceback.format_exc()}')
-        sys.exit(1)
-
-
-if __name__ == '__main__':
-    main()