The NLP Model Forge: Generate Model Code On Demand

Quantum Stat first came through with The Big Bad NLP Database, a collection of freely-accessible NLP datasets, curated from around the internet. It then released The Super Duper NLP Repo, which, at the time of introduction, provided centralized access to 100 freely-accessible NLP notebooks, curated from around the internet, and ready to launch in Colab with a single click. Now Quantum Stat is back with arguably its most ambitious NLP clearinghouse product yet.

The NLP Model Forge is here to help you create NLP models quickly and easily. As conveyed to me by Quantum Stat CEO Ricky Costa:

[The NLP Model Forge] allows users to generate code snippets from 1,400 NLP models curated from top NLP research companies such as Hugging Face Facebook DeepPavlov and AI2.

The general idea of The NLP Model Forge is that you are able to browse and search example model code by task type, which include sequence and token classification, question answering, text summarization, text generation, translation, and more. Each model has a number of attributes, which include:

• name (e.g. distilbert-base-uncased-finetuned-sst-2-english)
• source language
• type (e.g. DistilBERT)
• tags(e.g. pytorch, tf, rust, distilbert, text-classification, sst-2 dataset)
• labels (e.g. 0: NEGATIVE, 1: POSITIVE)

All of these model attributes, when taken together, provide a rich description and overview of realistic model expectations and implementation details. For example, it can be easily gleaned from the above example model's attributes that it is a DistilBERT text classification (sentiment analysis) model with implementations available in at least PyTorch and TensorFlow.

Once a desired model has been located, one can toggle the "load" switch in order to queue up the model code, after which the "get code" button can be used to load in the selected model code. See the image below for more details.

Once the code screen has been opened, the particular model's code can be viewed directly. Alternatively, a Colab notebook can be launched with a single click in order to have the model code snippet opened into an executable environment. See the image below for more details.

The NLP Model Forge seems best suited to individuals looking to bounce around and try out different models in their own projects, without needing to extensively research the ins and outs of each prior to giving them a test run. Once a suitable model has been settled upon, I could see the example code snippets being modified, extended, and personalized for maximum implementation effect. It would best be embraced as a fast iteration prototype assist than complete code for an end product, but, to be fair, nowhere is it stated that The NLP Model Forge is anything but this.

The NLP Model Forge is a promising project, and a worthy addition to Quantum Stat's NLP product repertoire. If you are looking for a centralized spot for browsing and searching model code snippets, or to get ideas for your next NLP project, it is definitely worth a visit.

 
Related:

• The Big Bad NLP Database: Access Nearly 300 Datasets
• The Super Duper NLP Repo: 100 Ready-to-Run Colab Notebooks
• Natural Language Processing Recipes: Best Practices and Examples