dynet

library2.1.2pypypiunverified

DyNet is the Dynamic Neural Network Toolkit, a C++ library with Python bindings for training neural networks, particularly focused on natural language processing. Current version 2.1.2 supports implicitly cast parameters as expressions, Python 3.8+, and advanced slicing. Release cadence is irregular.

ai-ml

pip install dynet

github ↗package ↗homepage ↗

INSTALL

IMPORT

SIG · DYNET

dynet

ai-mlpythonv2.1.2

Install

—

Import

—

Disk

—

Pass rate

0/ 10

Env Coverage0 / 10

glibc

3.9–3.13

musl

3.9–3.13

Install & Compatibility

Where this runs

tested against v? · pip install

Install × environment matrix

Each cell = how many times install + import succeeded across repeated harness runs. Partial = flaky.
glibc = Debian/Ubuntu slim · musl = Alpine Linux

musl

py 3.10–3.95 runs

build_error

glibc

py 3.10–3.95 runs

build_error

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

dynet

✓ import dynet as dy

✗ from dynet import *

Wildcard imports pollute namespace; convention is to import as dy.

Model

✓ from dynet import Model

✗ import dynet.Model

dynet.Model is not a submodule; directly import the class.

Minimal example: create a model, parameters, and perform a simple optimization loop.

import dynet as dy model = dy.Model() trainer = dy.SimpleSGDTrainer(model) pW = model.add_parameters((2, 2)) pb = model.add_parameters(2) x = dy.vecInput(2) y = dy.scalarInput(0) for epoch in range(5): loss = dy.pickneglogsoftmax(pW.expr() * x, y) loss.backward() trainer.update() print(f"Epoch {epoch}: loss = {loss.value()}")

Debug

Known issues

gotchaIn DyNet 2.1, parameters are implicitly cast to expressions. Do not call `dy.parameter(p)` or `p.expr()`; just use `p` directly.

fix

Use `p` instead of `p.expr()` or `dy.parameter(p)`.

affects: >=2.1

breakingDyNet 2.0 removed dependency on Boost. Model files saved with v1.x are incompatible. You must re-train or convert models.

fix

Retrain models with v2.0+ or use a conversion script (if available).

affects: >=2.0

gotchaAlways call `dy.renew_cg()` before creating new computation graph nodes in a loop, otherwise memory usage grows unbounded.

fix

Call `dy.renew_cg()` at the start of each training iteration.

affects: all

Upgrade

Version history

2.1.2latest on PyPI · released Oct 21, 2020

Audit

Dependencies

No dependency data recorded yet.

Agent activity

19 hits · last 30 days

node

OpenAI (training)

Resources

githubhttps://github.com/clab/dynet ↗

packagehttps://pypi.org/project/dynet/ ↗

homepagehttps://dynet.io ↗