bzamecnik

Installing CUDA 8.0 on PC with with GTX 980 Ti and Ubuntu 16.04

State as of 2017-07-31.

Recently I was installing CUDA on a Azure NC6 VM with Tesla K80 and later the same day I also upgraded my personal machine with GTX 980 Ti from Ubuntu 15.10 to 16.04 and completely reinstalled CUDA.

I had NVIDIA driver 361, CUDA 7.5 and cuDNN 4 or 5 and wanted CUDA 8.0

bzamecnik

import keras.backend as K
from keras.layers import Lambda
from keras.layers.merge import concatenate
from keras.models import Model
import keras.optimizers
from keras.optimizers import clip_norm, Optimizer
import tensorflow as tf

# this should be fairly ready
class DataParallelOptimizer(Optimizer):

bzamecnik

'''Trains a simple deep NN on the MNIST dataset.

Gets to 98.40% test accuracy after 20 epochs
(there is *a lot* of margin for parameter tuning).
2 seconds per epoch on a K520 GPU.
'''

from __future__ import print_function

import keras

bzamecnik

# Question: can we replicate https://github.com/rossumai/keras-multi-gpu/issues/2
# InvalidArgumentError: Shape [-1] has negative dimensions
#	 [[Node: replica_0_1/model_1_sample_weights = Placeholder[dtype=DT_FLOAT, shape=[?], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]

import tensorflow as tf

a = tf.placeholder(dtype=tf.int8, shape=(None, None))
b = tf.placeholder(dtype=tf.int8, shape=(None, None))
c = tf.add(a, b)

bzamecnik

Let's compile TensorFlow ourselves with AVX vector instruction to speed up inference on a CPU.

OSX:

https://www.tensorflow.org/install/install_sources#prepare_environment_for_mac_os tensorflow/tensorflow#12979

• fix checksum in tensorflow/workspace.bzl
• e5fdeee6b28cf6c38d61243adff06628baa434a22b5ebb7432d2a7fbabbdb13d but wanted 6d43b9d223ce09e5d4ce8b0060cb8a7513577a35a64c7e3dad10f0703bf3ad93
• comment the sha256 checksums tensorflow/tensorflow#6729

bzamecnik

              _              _          ___  
    ___   ___| |_ ___  _ __ (_)      .-'   `'. 
   / _ \ / __| __/ _ \| '_ \| |     /         \  
  | (_) | (__| || (_) | |_) | |     |         ;
   \___/ \___|\__\___/| .__/|_|     |         |           ___.--, 
     Raspbian         |_|  _.._     |0) ~ (0) |    _.---'`__.-( (_.
     GNU/Linux 7    __.--'`_.. '.__.\    '--. \_.-' ,.--'`     `""`
                   ( ,.--'`   ',__ /./;   ;, '.__.'`    __
                   _`) )  .---.__.' / |   |\   \__..--""  """--.,_
                  `---' .'.''-._.-'`_./  /\ '.  \ _.-~~~````~~~-._`-.__.'

bzamecnik

package com.github.bzamecnik.regex

import scala.concurrent.duration._

/**
  * Wraps String to make its usage time limited.
  *
  * After given time duration charAt() operation will start throwing a
  * TimeoutException.
  *

bzamecnik

=== Track 0
<message note_on channel=0 note=69 velocity=100 time=0>
<message note_off channel=0 note=69 velocity=0 time=522>
<message note_on channel=0 note=71 velocity=100 time=27>
<message note_off channel=0 note=71 velocity=0 time=27>
<message note_on channel=0 note=69 velocity=100 time=27>
<message note_off channel=0 note=69 velocity=0 time=27>
<message note_on channel=0 note=71 velocity=100 time=27>
<message note_off channel=0 note=71 velocity=0 time=137>
<message note_on channel=0 note=72 velocity=100 time=27>

bzamecnik

// list properties on an object - poor man's reflection
for (var prop in curScore) {
  console.log(prop)
}

//// properties of Ms:Score:
// addText
// appendMeasures
// appendPart
// composer

bzamecnik

# It's possible to serialize music21 stream to pickle.
# But for 1 kB of MIDI the pickle is 89 kB (!), after gzip only 8 kB.
# The benefit can be lower loading time compared to parsing MIDI
# at the cost of roughly 8x file size.

import music21
stream = music21.converter.parse("foo.mid")
pickled = music21.converter.freezeStr(stream)
with open("foo.pickle", "wb") as f:
    f.write(pickled)