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

Hmtl

Hmtl is an open-source Python implementation of a hierarchical multi-task NLP model that combines named entity recognition, entity mention detection, relation extraction and coreference resolution, with code for training, fine-tuning and evaluation.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Hmtl — 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 Hmtl when you want a Python codebase for experimenting with a hierarchical combination of semantic NLP tasks and can provide the configuration and datasets for your experiment.

Who it is for

  • NLP researchers studying shared representations across semantic tasks.
  • Developers adapting a multi-task model to a domain-specific combination of entity and relation tasks.

Documented uses

  • Train or fine-tune a hierarchical model combining entity recognition, mention detection, relation extraction and coreference resolution.
  • Evaluate learned embeddings and linguistic properties across the model hierarchy with SentEval.
Before you choose

Requirements & limits

The repository is public and lists an MIT license. The README specifies Python 3.6; Git LFS is needed to download pretrained model weights.

  • The implementation uses PyTorch and AllenNLP, with dependencies listed in requirements.txt.
  • Experiments are driven by JSON configuration files, and custom DatasetReaders may be needed.
  • Training datasets are not bundled and must be collected separately.
  • SentEval is an additional dependency specifically for evaluating embeddings.

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

Prepare and launch an HMTL training run

Set up the documented Python environment and start training from an experiment configuration.

Before you start

  • Python 3.6
  • The repository and its requirements.txt dependencies, including AllenNLP and PyTorch
  • A JSON configuration describing datasets, model and hyperparameters
  • Required datasets placed in the data/ folder; custom DatasetReaders may be needed

Steps

  1. Use ./script/machine_setup.sh to install Python 3.6, create a virtual environment and install the listed dependencies; adapt the script as needed.

  2. Prepare a JSON experiment configuration, using the examples in configs/ as a reference.

  3. Place the datasets expected by the configuration in data/ and define custom DatasetReaders if required.

  4. Run python train.py --config_file_path configs/hmtl_coref_conll.json --serialization_dir my_first_training.

Expected outcome

The documented command starts training from the selected configuration and assigns the experiment output to my_first_training.

Guide limitations

  • The training datasets are not included and must be collected separately; the README names OntoNotes 5.0, CoNLL2003 and ACE2005.
  • The setup script may need adaptation for a user's needs.
  • The repository gives a training example, not a guarantee that a particular environment or dataset will work unchanged.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
huggingface/hmtl · Official homepage title · retained in its source language
How we review sourcesSuggest a correction