dynet

Version:: 2.1.2
Category:: ai
Cluster:: Loki

Author / Distributor

Description

DyNet is a neural network library developed by Carnegie Mellon University and many others. It is written in C++ (with bindings in Python) and is designed to be efficient when run on either CPU or GPU, and to work well with networks that have dynamic structures that change for every training instance. For example, these kinds of networks are particularly important in natural language processing tasks, and DyNet has been used to build state-of-the-art systems for syntactic parsing, machine translation, morphological inflection, and many other application areas.

Documentation

$ python -m dynet.__main__ --help

usage: __main__.py [-h] [--dynet-seed SEED] [--dynet-mem MEM]
                   [--dynet-gpus GPUS] [--dynet-autobatch [LEVEL]]
                   [--dynet-devices DEVICES] [--dynet-viz]
                   [--dynet-weight-decay LAMBDA]
                   [--dynet-profiling] [--dynet-rand-seed SEED]
                   [--dynet-param-cgpu]
                   script.py ...

Common DyNet options:
  --dynet-seed SEED         Random seed
  --dynet-mem MEM           Memory to reserve (in MB)
  --dynet-gpus GPUS         Number of GPUs to use
  --dynet-autobatch [LEVEL] Enable dynamic auto-batching (0 = off, 1 = on)
  --dynet-devices DEVICES   Specific devices to use
  --dynet-viz               Enable graph visualization
  --dynet-weight-decay      Apply L2 weight regularization
  --dynet-profiling         Enable profiling
  --dynet-rand-seed         Seed for random initialization

Examples/Usage

Load the DyNet module:

$ module load ai/dynet-py37-cuda10.2-gcc8/2.1.2

Unload the module:

$ module unload ai/dynet-py37-cuda10.2-gcc8/2.1.2

Sample Python usage:

import dynet as dy

m = dy.ParameterCollection()
W = m.add_parameters((1, 2))
b = m.add_parameters(1)

x = dy.inputVector([0.5, 0.3])
y = dy.tanh(W * x + b)

print(y.value())

Installation

Source code is obtained from DyNet GitHub