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

Lm Spanish

Lm Spanish is a public repository of Spanish-language models and related NLP resources. It links to models including BNE RoBERTa and GPT-2 checkpoints, the multilingual Ǎguila-7B model, task-specific fine-tuned models, corpora, datasets, word embeddings and spaCy pipelines. Its README also includes Python examples for using BNE RoBERTa models in masked-token prediction.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Lm Spanish — 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 when selecting a Spanish NLP checkpoint or related resource and when you want a repository-level Python example for RoBERTa masked-token prediction.

Who it is for

  • Researchers and developers looking for Spanish NLP model and dataset resources.
  • Teams comparing Spanish-only and Spanish-Catalan-English model options.
  • Practitioners building experiments around masked-token prediction, named-entity recognition, part-of-speech tagging or question answering.

Documented uses

  • Find linked Spanish-language model checkpoints and related datasets.
  • Identify resources tuned for tasks such as POS tagging, NER and question answering.
  • Use the documented Transformers example to generate candidate tokens for a masked Spanish sentence.
  • Review the corpus scale and stated training-data context associated with selected Spanish models.
  • Locate word embeddings and domain-specific biomedical or legal language resources.
Before you choose

Requirements & limits

The repository is public on GitHub and identifies an Apache-2.0 license. Model and dataset resources are linked to their respective repositories or pages.

  • The repository is a resource collection, not a hosted inference service; individual models and datasets are linked elsewhere.
  • The listed models serve different purposes and language scopes: BNE RoBERTa is described for Spanish, Ǎguila-7B for Spanish, Catalan and English, and the Longformer entry for Catalan long-context masked-language modeling
  • The README’s executable example covers masked-token prediction with RoBERTa, not a general-purpose generation workflow.
  • The BNE corpus entry reports 570 GB and more than 135 billion tokens; the associated Spanish-model description specifies data compiled from National Library of Spain web crawls from 2009 to 2019.

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

Use a listed RoBERTa model for masked-token prediction

Load a BNE RoBERTa checkpoint and use the documented fill-mask pipeline on a Spanish sentence.

Before you start

  • Python
  • The transformers package and access to the named model repository, as used in the README example.

Steps

  1. Import AutoTokenizer, AutoModelForMaskedLM and FillMaskPipeline from transformers.

  2. Load the tokenizer and masked-language model for PlanTL-GOB-ES/roberta-base-bne or PlanTL-GOB-ES/roberta-large-bne.

  3. Set the model to evaluation mode and construct a FillMaskPipeline with the model and tokenizer.

  4. Pass a Spanish sentence containing the model’s token to the pipeline and inspect the returned token strings.

Expected outcome

A list of candidate tokens for the masked position in the supplied Spanish sentence.

Guide limitations

  • The example documents masked-token prediction for the two named RoBERTa checkpoints; it does not document a general text-generation workflow.
  • Model loading and execution depend on the linked checkpoint and local environment; the README does not specify hardware requirements.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
PlanTL-GOB-ES/lm-spanish · Official homepage title · retained in its source language
How we review sourcesSuggest a correction