OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Neuralconvo

Neuralconvo is a Lua/Torch implementation attempt of a neural conversational model based on sequence-to-sequence research. It uses two LSTM layers: one encodes an input sentence into a thought vector, and the other decodes it into a response. The documented experiment trains on the Cornell Movie-Dialogs Corpus.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Neuralconvo — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it as a code-level reference or experiment when you want to examine an LSTM-based seq2seq dialogue workflow in Torch and can account for the project's stated reproducibility issue.

Who it is for

  • Developers studying Lua and Torch implementations of neural dialogue models.
  • Researchers exploring sequence-to-sequence training on conversational text.

Documented uses

  • Set up the documented Torch and Lua dependencies and dialogue dataset.
  • Train a seq2seq dialogue model using the repository's training script.
  • Load the saved model and conduct a text conversation through the evaluation script.
Before you choose

Requirements & limits

  • The implementation uses Lua and Torch, with nn, rnn and penlight listed as additional dependencies.
  • Training data must be downloaded separately and placed in data/cornell_movie_dialogs.
  • The README documents CPU training as well as CUDA and OpenCL options, each with distinct setup requirements.
  • The author warns that the reported training results could no longer be reproduced after a dependency change.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Train the seq2seq model with CUDA

Train the model on 50,000 examples with a hidden size of 1,000 using CUDA.

Before you start

  • Torch command-line environment with th available.
  • CUDA support for the --cuda option.
  • The training data and dependencies required by train.lua.

Steps

  1. Run th train.lua --cuda --dataset 50000 --hiddenSize 1000.

  2. Allow the training process to run; the README reports a historical training time of three days on a GeForce GTX 780M.

Expected outcome

A trained model; the README describes sample conversations after 20 epochs with these settings.

Guide limitations

  • The README warns that the reported results can no longer be reproduced, apparently because a dependency changed.
  • The three-day training time is a historical report for a specific GPU, not a general runtime estimate.
Documentation 1 ↗
Source-based practical guide

Select a training compute mode

Choose CUDA, OpenCL, or CPU when invoking the documented training command.

Before you start

  • Torch command-line environment with th available.
  • The train.lua training script and its required data and dependencies.

Steps

  1. For CUDA, include --cuda in the training command.

  2. For OpenCL, use --opencl instead of --cuda.

  3. For CPU training, omit both --cuda and --opencl.

Expected outcome

The training command is configured for the selected compute mode.

Guide limitations

  • The README warns that its reported training results can no longer be reproduced, apparently due to a dependency change.
  • No further setup or compatibility details for these compute modes are provided.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗Guide documentation ↗
Source record details
Source availability
Active
Sources checked
7 November 2026
Original website title
macournoyer/neuralconvo · Official homepage title · retained in its source language
How we review sourcesSuggest a correction