fujiyuu75

pytorch_total_params = sum(p.numel() for p in model.parameters() if p.requires_grad)

fujiyuu75

def freeze_network(model):
    for name, p in model.named_parameters():
        if "l0" not in name:
            p.requires_grad = False
    return model

---

model = freeze_network(model)

fujiyuu75

class EarlyStopping:
    def __init__(self, patience=7, mode="max", delta=0.001):
        self.patience = patience
        self.counter = 0
        self.mode = mode
        self.best_score = None
        self.early_stop = False
        self.delta = delta
        if self.mode == "min":
            self.val_score = np.Inf

fujiyuu75

import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
from skimage.measure import compare_ssim
import imutils
import cv2
from tqdm import tqdm

TRAIN_VIDEOS='../input/deepfake-detection-challenge/train_sample_videos'
TEST_VIDEOS='../input/deepfake-detection-challenge/test_videos'
ALL_VIDEOS='../input/dfdc_train_all'

fujiyuu75

### Model Visualizations

def normalize_channels(img):
    _min, _max = img.min(axis=(0,1)), img.std(axis=(0,1))
    img = (img - _min) / (_max - _min)
    return img

def plot_first_kernels(weight):
    ''' plot first filters of a model '''
    with torch.no_grad():

fujiyuu75

def seed_everything(seed):
    random.seed(seed)
    os.environ['PYTHONHASHSEED'] = str(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    torch.backends.cudnn.deterministic = True

fujiyuu75

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

fujiyuu75

## Memory optimization

# Original code from https://www.kaggle.com/gemartin/load-data-reduce-memory-usage by @gemartin
# Modified to support timestamp type, categorical type
# Modified to add option to use float16

from pandas.api.types import is_datetime64_any_dtype as is_datetime
from pandas.api.types import is_categorical_dtype

def reduce_mem_usage(df, use_float16=False):

fujiyuu75

plt.figure(figsize=(10,10))
for i in range(25):
    plt.subplot(5,5,i+1)
    plt.xticks([])
    plt.yticks([])
    plt.grid(False)
    plt.imshow(train_images[i], cmap=plt.cm.binary)
    plt.xlabel(class_names[train_labels[i]])
plt.show()

fujiyuu75

plt.figure()
plt.imshow(train_images[0])
plt.colorbar()
plt.grid(False)
plt.show()