mkolod

#include <chrono>
#include <iostream>
#include <vector>
#include <thread>

__global__ void do_nothing(int time_us, int clock_rate) {
  clock_t start = clock64();
  clock_t end;
  for (;;) {
    end = clock64();

mkolod

#include <csignal>
#include <cstdlib>
#include <fstream>
#include <iostream>
#include <unistd.h>
#include <limits.h>

#include <iostream>
#include <sstream>
#include <stdexcept>

mkolod

ImageNet validation set fix:

0. The training set is organized in directories, with each directory matching a class, e.g. "n01751748" matching "sea snake." However, the valset is a flat dir of JPEGs. The ImageNet labels provided in the devkit for the validation set (ILSVRC2012_validation_ground_truth.txt) are not consistent with the ordering used by PyTorch/TF/Keras/MXNet/Caffe, etc. for pre-trained models. For example, in the the above ground truth label file, "sea snake" is 490, but in PyTorch/TF, it's 65.
   Proof:
   • ImageNet validation labels: https://gist.githubusercontent.com/aaronpolhamus/964a4411c0906315deb9f4a3723aac57/raw/aa66dd9dbf6b56649fa3fab83659b2acbf3cbfd1/map_clsloc.txt
   • DL framework labels: https://gist.githubusercontent.com/aaronpolhamus/964a4411c0906315deb9f4a3723aac57/raw/aa66dd9dbf6b56649fa3fab83659b2acbf3cbfd1/map_clsloc.txt
1. Untar the valset file, you will get a flat dir of JPEGs.
2. Pull in the unflattening script into the directory where the val images were unpa

mkolod

import torch.cuda
import torch.cuda.memory as cumem
import sys
import ctypes as C

GB = 1 << 30

def get_cuda_memory():
    handle = C.cdll.LoadLibrary("libcudart.so")
    free, total = C.c_long(), C.c_long()

mkolod

" Use Vim settings, rather then Vi settings (much better!).
" This must be first, because it changes other options as a side effect.
set nocompatible

" TODO: this may not be in the correct place. It is intended to allow overriding <Leader>.
" source ~/.vimrc.before if it exists.
if filereadable(expand("~/.vimrc.before"))
  source ~/.vimrc.before
endif

mkolod

'''
Memory profiling utilities
'''
import gc
import inspect
import linecache
import os.path
import sys
import time
import threading

mkolod

import onnx
import sys

name = sys.argv[1]
model = onnx.load(name)
onnx.checker.check_model(model)
print(onnx.helper.printable_graph(model.graph))

mkolod

RUN wget http://developer.download.nvidia.com/compute/cuda/10.2/Prod/local_installers/cuda_10.2.89_440.33.01_linux.run
RUN chmod +x cuda_10.2.89_440.33.01_linux.run && \
    ./cuda_10.2.89_440.33.01_linux.run --silent --toolkit
    
# After installing CUDA, you need to add the env vars below, otherwise you'll get CUDA runtime version/driver
# version issue, even if you're using nvidia-docker.
    
ENV LIBRARY_PATH=/usr/local/cuda/lib64/stubs
ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility

mkolod

# precondition: d is power of 2
def mod_pow2(n, d): 
    return ( n & (d-1) ) 
 
print(n, "mod" , d , "is", mod_pow2(6, 4)) 
    

mkolod

#include <iostream>
#include <sys/mman.h>

template<typename T>
T* pinned_malloc(size_t num_elem) {

    size_t num_bytes = num_elem * sizeof(T);
    T *ptr = static_cast<T *>(malloc(num_bytes));

    if (ptr == nullptr) {