samsung/veles

Distributed machine learning platform

Veles

Distributed platform for rapid Deep learning application development

Consists of:

• Platform - https://github.com/Samsung/veles

• Znicz Plugin - Neural Network engine

• Mastodon - Veles <-> Java bridge for Hadoop etc.

• SoundFeatureExtraction - audio feature extraction library

Home page: https://velesnet.ml

Named after https://en.wikipedia.org/wiki/Veles_(god)

Released under Apache 2.0 license. Copyright © Samsung Electronics Co., Ltd., 2013-2015.

veles's Issues

multivariate time series prediction

hi all,
just wondering if there is a multivariate time series prediction in this frame work.

Thanks,
Andrew

Fix ImagenetAE fine tuning and loader

Data is in a new format after refactoring preparation_imagenet, so we need to fix loader in ImagenetAE. And plotters does not work at fine tuning stage.

Add Imagenet forward propagation

Samsung/veles.znicz#35

Fix misprint in copyright

Refactor InputJoiner

Add link_inputs method to InputJoiner class, add input_0, input_1, ..., offset_0, offset_1, ..., length_0, length_1, ... properties for each of the input.

Fix background types assert in image loader

Does veles supports opencl 1.1 EP

Hello,

we are looking for a deep learning platform to train against a vivante gc2000 gpu from the imx6q board which supports only opencl 1.1 Embedded Profile.

Does Veles supports that limited version of opencl?

Thanks

Update documentation

Include OpenCL driver version in the cache

When driver version changes, binaries become invalid and result in errors. We should include it into the cache.

veles documentation is unavailable

Hello,

on veles website, the "docs" tab leads to a 404 : https://velesnet.ml/docs
Same problem with https://velesnet.ml/jenkins

I just wanted to warn you
Many thanks !

Check input/output data types

Add check of data types comparability (float/double) in initialize

Add script to copy samples from veles to user's directory

Support global metrics: ROC, AUC, Lift chart, r-squared.

Fix incorrect LinkableAttribute behaviour when attribute already contained in object's __dict__

Reanimate libVeles

libVeles needs emergency amending to support the latest platform.

Fix automatic uploading of Models to Forge

Fix bug with multithreaded OpenCL matrix multiplication without clBLAS

Fix Gulp build

Recent plugin updates ruined the package build: https://velesnet.ml/jenkins/job/VELES_Package_Release/84/

We need to set exact dependency versions.

Check and fix bugs in command line Frontend GUI

1. Fix paths to correct work of Frontend on developer's Veles version

lpodoynitsina@lpodoynitsinau64:~/Projects/Veles$ python3 -m veles --frontend

Processing arguments...
Writing /export/home/lpodoynitsina/Projects/Veles/web/dist/frontend.html...
Process Process-1:
Traceback (most recent call last):
File "/usr/lib/python3.4/multiprocessing/process.py", line 254, in _bootstrap
self.run()
File "/usr/lib/python3.4/multiprocessing/process.py", line 93, in run
self._target(_self._args, *_self._kwargs)
File "/export/home/lpodoynitsina/Projects/Veles/veles/main.py", line 282, in _open_frontend_process
main()
File "/export/home/lpodoynitsina/Projects/Veles/veles/scripts/generate_frontend.py", line 83, in main
sout = loader.load("frontend.html").generate(
File "/usr/local/lib/python3.4/dist-packages/tornado/template.py", line 343, in load
self.templates[name] = self._create_template(name)
File "/usr/local/lib/python3.4/dist-packages/tornado/template.py", line 370, in _create_template
with open(path, "rb") as f:
FileNotFoundError: [Errno 2] No such file or directory: '/export/home/lpodoynitsina/Projects/Veles/web/dist/templates/frontend.html'

Save best configuration into file after genetic optimization

Add BLAS-style alpha and beta for updating backpropagated error

Required changes for veles.znicz's same issue.

Add example of genetic optimization

Fix tests for 0.8.2 release #10

Samsung/veles.znicz#10

GUI for creating workflow

Get rid of "snapshot" keyword argument in Unit.initialize()

Engine can be smart enough to figure it out itself.

Fix automatically downloading of files. Set correct paths for data

Samsung/veles.znicz#19

Split documentation between Veles and Znicz

Samsung/veles.znicz#37

Resurrect Python 2 compatibility

Fix possible incompatibilities with Python 2.7 after the transfer to GitHub infra.

init script issue

The init script produces:

fatal: reference is not a tree: 4e8093554080bde69bc66ca15f6fe63de573ef8a
Unable to checkout '4e8093554080bde69bc66ca15f6fe63de573ef8a' in submodule path 'veles/znicz'
Failed to checkout submodules, retrying...
fatal: reference is not a tree: 4e8093554080bde69bc66ca15f6fe63de573ef8a
Unable to checkout '4e8093554080bde69bc66ca15f6fe63de573ef8a' in submodule path 'veles/znicz'
Failed to checkout submodules, retrying...
fatal: reference is not a tree: 4e8093554080bde69bc66ca15f6fe63de573ef8a
Unable to checkout '4e8093554080bde69bc66ca15f6fe63de573ef8a' in submodule path 
...

during a fresh install.

Refactor gradient update in master-slave mode

Now, gradient update in master-slave mode is the simplest with manually applying the moment.
It is required to refactor in such way, that gradient update will call the same existing methods and without copy-paste in master-slave mode.

Installation error

paramico versionis not correct

Add support for Ubuntu 15.04

Fix genetic optimization in master-slave mode

Launch master first:
python3 -m veles -p '' -s --optimize 25 -l 0.0.0.0:10000 veles/samples/genetics.py -
In another console launch slave:
python3 -m veles -p '' -s --optimize 25 -m localhost:10000 veles/samples/genetics.py -

Fix Veles installation Ubuntu 15.04

Command python setup.py egg_info failed with error code 1 in /tmp/pip-build-hv1t6anb/pygit2
Storing debug log for failure in /home/lyubov/.pip/pip.log
dpkg: error processing package python3-veles (--configure):
subprocess installed post-installation script returned error exit status 1
Setting up odbcinst (2.3.1-3) ...
Processing triggers for libc-bin (2.21-0ubuntu4) ...
Processing triggers for ureadahead (0.100.0-19) ...
Errors were encountered while processing:
python3-veles
E: Sub-process /usr/bin/dpkg returned an error code (1)

description cause mis-understanding.

Veles actually is a deep learning(DNN) application. But its description says "Distributed machine learning platform". Deep Learning is a subset of Machine Learning, so it should be "Distributed deep learning platform". Thanks.

Fix forward extraction. Add forward extraction example

Samsung/veles.znicz#15

Add the way to extract dereferenced, real unit objects from links_from/to

Unit.links_from and Unit.links_to may contain weak references in order to enable garbage collection. It may confuse users who try to read linked units.

Add automatically disabling of snapshotting in genetic mode

Need to disable snapshotting in genetic mode because large number of snapshots is overflowing memory.

Change suffix of snapshot

Remove "none" suffix if there is no samples in some sets (test, validation, train). Add "test" results to the suffix

Fix registration on VelesForge

ERROR:tornado.application:Future <tornado.concurrent.Future object at 0x7fa0a9283518> exception was never retrieved: tornado.iostream.StreamClosedError
ERROR:tornado_smtpclient.client:
Traceback (most recent call last):
File "/usr/local/lib/python3.4/dist-packages/tornado_smtpclient/client.py", line 101, in getreply
response = yield self.stream.read_until(CRLF)
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 870, in run
value = future.result()
File "/usr/local/lib/python3.4/dist-packages/tornado/concurrent.py", line 215, in result
raise_exc_info(self._exc_info)
File "", line 3, in raise_exc_info
tornado.iostream.StreamClosedError
ERROR:tornado.application:Uncaught exception GET /forge/service?query=register&email=[email protected] (127.0.0.1)
HTTPServerRequest(protocol='http', host='velesnet.ml', method='GET', uri='/forge/service?query=register&email=[email protected]', version='HTTP/1.0', remote_ip='127.0.0.1', headers={'Host': 'velesnet.ml', 'Cookie': '__utma=223305372.1977894979.1433147698.1447756335.1447842103.14; __utmc=223305372; __utmz=223305372.1447677400.9.2.utmcsr=t.co|utmccn=(referral)|utmcmd=referral|utmcct=/Ubbg2QnW6l; ga=GA1.2.1977894979.1433147698; gat=1', 'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,/;q=0.8', 'Accept-Encoding': 'gzip, deflate, sdch', 'Accept-Language': 'en-US,en;q=0.8', 'X-Forwarded-For': '212.44.150.238', 'Connection': 'close', 'Upgrade-Insecure-Requests': '1', 'User-Agent': 'Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/45.0.2454.101 Chrome/45.0.2454.101 Safari/537.36'})
Traceback (most recent call last):
File "/usr/local/lib/python3.4/dist-packages/tornado_smtpclient/client.py", line 101, in getreply
response = yield self.stream.read_until(CRLF)
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 870, in run
value = future.result()
File "/usr/local/lib/python3.4/dist-packages/tornado/concurrent.py", line 215, in result
raise_exc_info(self._exc_info)
File "", line 3, in raise_exc_info
tornado.iostream.StreamClosedError
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/usr/local/lib/python3.4/dist-packages/tornado/web.py", line 1369, in _stack_context_handle_exception
raise_exc_info((type, value, traceback))
File "", line 3, in raise_exc_info
File "/usr/local/lib/python3.4/dist-packages/tornado/stack_context.py", line 314, in wrapped
ret = fn(_args, *_kwargs)
File "/usr/local/lib/python3.4/dist-packages/tornado/web.py", line 1581, in future_complete
f.result()
File "/usr/local/lib/python3.4/dist-packages/tornado/concurrent.py", line 215, in result
raise_exc_info(self._exc_info)
File "", line 3, in raise_exc_info
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 876, in run
yielded = self.gen.throw(_exc_info)
File "/usr/lib/python3/dist-packages/veles/forge/forge_server.py", line 243, in get
yield self.handlersself.get_argument("query")
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 870, in run
value = future.result()
File "/usr/local/lib/python3.4/dist-packages/tornado/concurrent.py", line 215, in result
raise_exc_info(self._exc_info)
File "", line 3, in raise_exc_info
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 876, in run
yielded = self.gen.throw(_exc_info)
File "/usr/lib/python3/dist-packages/veles/forge/forge_server.py", line 163, in handle_register
smtp = yield self.server.smtp()
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 870, in run
value = future.result()
File "/usr/local/lib/python3.4/dist-packages/tornado/concurrent.py", line 215, in result
raise_exc_info(self._exc_info)
File "", line 3, in raise_exc_info
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 876, in run
yielded = self.gen.throw(_exc_info)
File "/usr/lib/python3/dist-packages/veles/forge/forge_server.py", line 641, in smtp
yield smtp.connect(self.smtp_host, int(self.smtp_port))
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 870, in run
value = future.result()
File "/usr/local/lib/python3.4/dist-packages/tornado/concurrent.py", line 215, in result
raise_exc_info(self._exc_info)
File "", line 3, in raise_exc_info
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 876, in run
yielded = self.gen.throw(_exc_info)
File "/usr/local/lib/python3.4/dist-packages/tornado_smtpclient/client.py", line 136, in connect
(code, msg) = yield self.getreply()
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 870, in run
value = future.result()
File "/usr/local/lib/python3.4/dist-packages/tornado/concurrent.py", line 215, in result
raise_exc_info(self._exc_info)
File "", line 3, in raise_exc_info
File "/usr/local/lib/python3.4/dist-packages/tornado/gen.py", line 876, in run
yielded = self.gen.throw(*exc_info)
File "/usr/local/lib/python3.4/dist-packages/tornado_smtpclient/client.py", line 105, in getreply
raise smtplib.SMTPServerDisconnected("Connection unexpectedly closed")
smtplib.SMTPServerDisconnected: Connection unexpectedly closed
ERROR:tornado.access:500 GET /forge/service?query=register&email=[email protected] (127.0.0.1) 127300.53ms

Add logging info in Snapshotter

Need to add logging info when Snapshotter is reading files

Implement new loss functions

Need to implement loss functions:

1. Sigmoid Cross-Entropy
2. Log Softmax

Remove music genre related files

Fix runing veles from python2

lpodoynitsina@lpodoynitsinau64:~/Projects/Veles$ python
Python 2.7.6 (default, Jun 22 2015, 17:58:13)
[GCC 4.8.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.

import veles
veles/init.py:67: UserWarning: Cannot expand variables generated by Git, setting them to None
warn("Cannot expand variables generated by Git, setting them to None")
kwargs={"dry_run": "init", "snapshot": file_name, "stealth": True}
Traceback (most recent call last):
File "", line 1, in
NameError: name 'file_name' is not defined
kwargs={"dry_run": "init", "stealth": True}
path_to_model = "veles/znicz/samples/MNIST/mnist.py"
launcher = veles(path_to_model, **kwargs)
Traceback (most recent call last):
File "", line 1, in
File "veles/init.py", line 179, in call
Main = import_module("veles.main").Main
TypeError: 'NoneType' object is not callable

Improve garbage collection in tests

Fix manhole option

When manhole is closing, spinner tries to write in closed stdout

Import caffe/torch models

Is it possible? Do you have a special tool for converting Nets?

Add automatically saving plotters to pdf after training

It is necessary to add the ability to save the results in PDF after the trainings, in any case. This is particularly necessary for a long time training models (for example, imagenet)

Add method of changing one Unit to another in Workflow with saving all data and control links

Fix MinibachesSaver-Loader

See YaleFaces example: yale_faces.py workflow with Minibatches Loader. And yale_faces_preprocessing.py workflow with Minibatches Saver. Currently Saver saves different data each time. And all preprocessing-loader-model process with MinibachesSaver-Loader gives different results, then with FullBatchAutoLabelImageLoader