amogh112

Command to run:
NCCL_DEBUG=WARN CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 ./scripts/dist_run_single.sh 6 contrastive_pretrain/train_end2end.py ./cfgs/contrastive_pretrain/base_prec_random_movienet_images_4x16G_fp32.yaml ./checkpoints_debug2 | tee debug2.txt

Namespace(cfg='./cfgs/contrastive_pretrain/base_prec_random_movienet_images_4x16G_fp32.yaml', cudnn_off=False, dist=True, do_test=False, log_dir='./checkpoints_debug2/./output/vl-bert/contrastive_random_images/base_prec_random_movienet_images_4x16G_fp32/train_train/tensorboard_logs', model_dir='./checkpoints_debug2', slurm=False)
Namespace(cfg='./cfgs/contrastive_pretrain/base_prec_random_movienet_images_4x16G_fp32.yaml', cudnn_off=False, dist=True, do_test=False, log_dir='./checkpoints_debug2/./output/vl-bert/contrastive_random_images/base_prec_random_movienet_images_4x16G_fp32/train_train/tensorboard_logs', model_dir='./checkpoints_debug2', slurm=False)
Namespace(cfg='./cfgs/contrastive_pretrain/base_prec_random_movienet_images_4x16G_fp32.yaml', cudnn_off=Fa

amogh112

import tqdm
from multiprocessing import Pool


def process_movie(movie):
    print("Processing movie : ", movie)
    path_movie_videos = os.path.join(path_all_movies, movie)
    list_avi_files = os.listdir(path_movie_videos)

    count=1

amogh112

def print_gc_tensors():

    dict_tensorsize_to_count = defaultdict(int) # Tensor size to number
    dict_device_to_tensors = defaultdict(lambda:  defaultdict(int)) # Device to tensor count
    dict_device_memory =  defaultdict(int) # Device to tensor memory

    total_count = 0
    for obj in gc.get_objects():
        try:
            if torch.is_tensor(obj) or (hasattr(obj, 'data') and torch.is_tensor(obj.data)):

amogh112

class VLBertClassifier(VLBert):

    def __init__(self, cfg, args, tok, num_layers, num_outputs, hidden_units=1024, dim_mlp=384):
        super(VLBertClassifier, self).__init__(cfg, args, tok)

        if num_layers == 2:
            self.final_mlp = torch.nn.Sequential(
                        torch.nn.Dropout(0.1, inplace=False),
                        torch.nn.Linear(dim_mlp, hidden_units),
                        torch.nn.ReLU(inplace=True),

amogh112

import time
from collections import defaultdict
import os
import gc
import torch
from tqdm import trange, tqdm

import masker
import tests
import utils

amogh112

import numpy as np


def generateData(a, b, num_samples):
    data_x = np.random.uniform(a, b, num_samples)
    data_y = np.random.uniform(a, b, num_samples)
    data = np.array(list(zip(data_x, data_y)))
    labels = np.array([-1 if datap[0] <= datap[1] else 1 for datap in data])
    return data, labels

amogh112

import cv2
import numpy as np
import time

drawing = False # true if mouse is pressed
ix,iy = -1,-1

# mouse callback function
def draw_circle(event,x,y,flags,param):
    global ix,iy,drawing

amogh112

import tensorflow as tf

def load_graph(frozen_graph_filename):
    # We load the protobuf file from the disk and parse it to retrieve the 
    # unserialized graph_def
    with tf.gfile.GFile(frozen_graph_filename, "rb") as f:
        graph_def = tf.GraphDef()
        graph_def.ParseFromString(f.read())

    # Then, we import the graph_def into a new Graph and returns it 

amogh112

#!/usr/bin/env python

import rospy
import sys
from std_msgs.msg import String, Header
from sensor_msgs.msg import CameraInfo
from sensor_msgs.msg import Image

class publishTF(object):

amogh112

#!/usr/bin/env python

import rospy
from std_msgs.msg import String
from sensor_msgs.msg import NavSatFix

class publishGPS(object):

	def __init__(self):
		rospy.loginfo("Initialising GPS publishing")