ilkarman

import argparse
import os
import shutil
import time
import socket
import multiprocessing
import numpy as np
import torch
import torch.nn as nn
import torch.nn.parallel

ilkarman

Step 36/51 : RUN git clone --recursive https://github.com/pytorch/vision  && cd vision  && git submodule sync && git submodule update --init --recursive  && export TORCH_CUDA_ARCH_LIST="3.5 5.2 6.0 6.1 7.0+PTX"  && export TORCH_NVCC_FLAGS="-Xfatbin -compress-all"  && export CMAKE_PREFIX_PATH="$(dirname $(which conda))/../"  && python setup.py clean  && python setup.py install
 ---> Running in 2ee1554613dc
Cloning into 'vision'...
No CUDA runtime is found, using CUDA_HOME='/usr/local/cuda'
/opt/conda/lib/python3.6/site-packages/torch/cuda/__init__.py:52: UserWarning: CUDA initialization: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx (Triggered internally at  ../c10/cuda/CUDAFunctions.cpp:100.)
  return torch._C._cuda_getDeviceCount() > 0
Building wheel torchvision-0.8.0a0+588f7ae
PNG found: True
libpng version: 1.2.54

ilkarman

Step 41/56 : RUN cd vision  && . /opt/conda/bin/activate  && export TORCH_CUDA_ARCH_LIST="3.7;6.1;7.0;7.5"  && export FORCE_CUDA=$torchvision_cuda  && python setup.py install
 ---> Running in ff2dd49d614a
No CUDA runtime is found, using CUDA_HOME='/usr/local/cuda'
Building wheel torchvision-0.7.0a0+78ed10c
running install
running bdist_egg
running egg_info
creating torchvision.egg-info
writing torchvision.egg-info/PKG-INFO

ilkarman

# Kill all processes in nvidia-smi
kill -9 $(nvidia-smi | sed -n 's/|\s*[0-9]*\s*\([0-9]*\)\s*.*/\1/p' | sort | uniq | sed '/^$/d'

# Resize param for ffmpeg: 171 width, 128 height & pad (without stretching)
-vf "scale=iw*min(171/iw\,128/ih):ih*min(171/iw\,128/ih),pad=171:128:(171-iw)/2:(128-ih)/2"

# Stop/remove all docker containers
docker stop $(docker ps -a -q)
docker rm $(docker ps -a -q)

ilkarman

import requests
import pandas as pd
LIM = 3000
OFFSET = 1000
all_papers = []
for offs in range(0,LIM,OFFSET):
    papers = requests.get(
    "https://openreview.net/notes?invitation=ICLR.cc%2F2020%2FConference%2F-%2FBlind_Submission&details=replyCount%2C"+
        "original&includeCount=true&offset={0}&limit={1}".format(offs, LIM))
    rows = [

ilkarman

import torch
import torchvision
import torch.utils.data
import random
import numpy as np
from torch.utils.data import TensorDataset

# https://github.com/galatolofederico/pytorch-balanced-batch/blob/master/sampler.py
class BalancedBatchSampler(torch.utils.data.sampler.Sampler):
    def __init__(self, dataset, labels=None):

ilkarman

import argparse
import logging
import os
from os import path
import numpy as np
import pandas as pd
import multiprocessing
import random
from toolz import pipe
from timer import Timer

ilkarman

    biases <- lapply(seq_along(listb), function(idx){
    r <- listb[[idx]]
    matrix(rnorm(n=r), nrow=r, ncol=1)
    })

    weights <- lapply(seq_along(listb), function(idx){
    c <- listw[[idx]]
    r <- listb[[idx]]
    matrix(rnorm(n=r*c), nrow=r, ncol=c)
    })

ilkarman

square_stack_lst_of_matricies <- function(lst)
{
    sqr_size <- sqrt(length(lst))
    # Stack vertically
    cols <- do.call(cbind, lst)
    # Split to another dim
    dim(cols) <- c(dim(filter_map[[1]])[[1]],
                   dim(filter_map[[1]])[[1]]*sqr_size,
                   sqr_size)
    # Stack horizontally

ilkarman

 SGD <- function(training_data, epochs, mini_batch_size, lr, C, sizes, num_layers, biases, weights,
                 verbose=FALSE, validation_data)
 {
   # Every epoch
   for (j in 1:epochs){
     # Stochastic mini-batch (shuffle data)
     training_data <- sample(training_data)
     # Partition set into mini-batches
     mini_batches <- split(training_data, 
                           ceiling(seq_along(training_data)/mini_batch_size))