innerlee

from:
https://github.com/tensorflow/tensorflow/blob/9e8a4938c06301cc3afa4cd3a3c6c2279e81c98f/tensorflow/core/ops/training_ops.cc#L443-L465

Update '*var' according to the RMSProp algorithm.
Note that in dense implement of this algorithm, ms and mom will 
update even if the grad is zero, but in this sparse implement, ms 
and mom will not update in iterations the grad is zero.

mean_square = decay * mean_square + (1-decay) * gradient ** 2
Delta = learning_rate * gradient / sqrt(mean_square + epsilon)

innerlee

sudo lsblk -f

innerlee

cat sth > /dev/null

innerlee

sudo addgroup imagenet --gid 1111
sudo adduser yxge --uid 1107
sudo adduser yxge imagenet

innerlee

sudo mkdir -p /media/ramdisk
sudo mount -t tmpfs -o size=2048M tmpfs /media/ramdisk

innerlee

find DIR_NAME -type f | wc -l

innerlee

@everywhere image_folder    = "/home/gv6/openimages/train_images/"
@everywhere tmp_folder      = "/media/ramdisk"
@everywhere save_folder     = "/home/lizz/openimage"
@everywhere missing_folder  = joinpath(save_folder, "missing")

@everywhere long, short, longest, shortest = [1024, 512, 8192, 128]

@everywhere function process(line)

innerlee

#!/usr/bin/python
# -*- coding: utf-8 -*-

# activation value

import sys
sys.path.append('/home/lizz/parrots/parrots/python')
from pyparrots import dnn
import numpy as np
from numpy import linalg as alg

innerlee

cat labels.csv | cut -c1-16 | sort | uniq | wc -l

innerlee

using HDF5
using JLD
using LIBLINEAR

println("> svm on open image")

# read features (2048,165659)
features = h5read("data/grand5_feature.h5", "global_pool")[1, 1, :, :]
ids = readdlm("data/redis_val_list_with_id.txt")[:, 1]