Sergei Mironov sergei-mironov

https://www.tensorflow.org/lite
- Small version of TF, targets mobile devices on Android and IoT
https://www.tensorflow.org/performance/xla/
- XLA (Accelerated Linear Algebra) is a domain-specific compiler for linear algebra that optimizes TensorFlow computations.

Responsibilities

Research on the related technologies of formal modeling and formal verification in operating systems, embedded hardware and software, and model-driven development tools
Responsible for exploration of cutting-edge technologies and cutting-edge solutions in the field of formal verification
Responsible for formal verification project operation and implementation
Participate in the formalization of relevant project practices in the field of operating systems, complete the research, design, coding, debugging, etc.

Народ активно работает над задачей генерации картинки по тексту
Особенно активно работают китайские исследователи
Но картинки пока получаются маленькие - 256x256 пикселей, хотя вроде бы красиые.
И на узкие темы. Например - только о птичках.
Чтобы сеть овладела вопросом, нужно примерно 10 тыс. картинок с описаниями.
Для работы над приложением задачи нужно:

Собрать и разметить датасет. Наверное, это самое основное и важное. Пример данных см. по ссылки ниже, в "датасеты". Размечальщиков картинок можно нанять, см. MTurk.

	attention input keep probability = 0.6
	AttentionCellWrapper added!
	W0922 22:44:48.935649 140431404037952 deprecation_wrapper.py:119] From /workspace/mironov/nl2bash/encoder_decoder/framework.py:200: The name tf.train.AdamOptimizer is deprecated. Please use tf.compat.v1.train.AdamOptimizer instead.

	W0922 22:44:58.016724 140431404037952 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/clip_ops.py:286: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.
	Instructions for updating:
	Use tf.where in 2.0, which has the same broadcast rule as np.where
	W0922 22:45:20.083132 140431404037952 deprecation_wrapper.py:119] From /workspace/mironov/nl2bash/encoder_decoder/framework.py:223: The name tf.train.Saver is deprecated. Please use tf.compat.v1.train.Saver instead.

	Making model_dir...

	#!/bin/sh

	RELAY="$1"
	PORT="$2"

	echo -n 'nat-hole-punch' \| socat STDIN "UDP-SENDTO:$RELAY:$PORT,sourceport=$PORT"
	mosh-server new -p "$PORT" \| sed -n 's/MOSH CONNECT [0-9]\+ \(.*\)$/\1/g p'

	#!/bin/sh

	MOSH_SSH_NAME=$1
	MOSH_RELAY=$2
	PORT_A=34730
	PORT_B=34731

	MOSH_KEY=$(ssh "$MOSH_SSH_NAME" "mosh-nat-server.sh \"$MOSH_RELAY\" \"$PORT_A\"")
	MOSH_KEY="$MOSH_KEY" mosh-client "$MOSH_RELAY" "$PORT_B"

	# This file defines Nix environment for building Ergo and Akylos GPU miner. To
	# enter the environmant:
	# 1. Install [Nix](https://nixos.org/nix)
	# package manager (or use NixOS Linux distro which is base on it).
	# 2. Save this file to current directory and run `nix-shell`. Nix will
	# download all the dependencies and open the shell where commands like
	# `sbt` or `nvcc` just work.
	#
	# Note on syntax: The first line `{ pkgs ? import <nixpkgs> {} } :` refera to
	# an instance of [nixpkgs](https://github.com/NixOS/nixpkgs) repository.

	# For the list of available plugins see
	# nixpkgs/pkgs/misc/vim-plugins/generated.nix
	#
	# Example of working config
	# nixcfg/src/pkgs/myvim.nix
	#
	{ pkgs ? import <nixpkgs> {} } :
	with pkgs;

	vim_configurable.customize {