You can now publish workflows on Clarifai Community. Workflows make it easy to connect models of different kinds so that you can perform complex operations on your data and build solutions that target your specific business needs. The nodes in your Mesh Workflows can accept data from other models, just select a model that works with the types of inputs and outputs that you need.
Clarifai now supports two-factor authentication (2FA) in New Portal! Having a second form of identification decreases the chance of a hacker gaining access to corporate devices or other sensitive information. Two-Factor authentication can also help reduce time-consuming password resets, and 2FA provides a safe way for users to reset their passwords. The outcome for businesses is increased employee productivity.
Clarifai is releasing a new model that leverages a “simple contrastive learning framework” (SimCSE) that works with unlabeled and labeled data. Unsupervised SimCSE simply takes an input sentence and predicts itself in a contrastive learning framework, with only standard dropout as noise. Our supervised SimCSE incorporates annotated pairs from NLI datasets into contrastive learning by using entailment pairs as positives and contradiction pairs as hard negatives. The contrastive learning objective regularizes pre-trained embeddings’ anisotropic space to be more uniform, and it better aligns positive pairs when supervised signals are available.
You can now complete labeling assignments directly in New Portal! Annotations (also called "labels") are how we "teach" machines to recognize concepts. When you want to create a custom model for your business, you do this by training this model to recognize the concepts that you have annotated on your sample data. Clarifai provides fully featured labeling tools for unstructured data.
You can now view summary statistics for Dataset-Versions for a particular Dataset. Track and view different dataset versions helps you experiment with your models and improve them iteratively. Summary statistics provide essential information about your dataset versions so that you can make informed decisions and work effectively as you iterate on your models.
Clarifai can now import Text-Embedder models from Huggingface. Text embeddings are critical to enabling “quick training” or “transfer learning”. Model embeddings can be used to train new models with few training examples quickly.
Clarifai can now import a Visual-Detector from MMDetection. This new capability provides a consistent, easy-to-understand solution for maintenance and future integration with MMDetection.
Clarifai is always looking for ways to accelerate and simplify the process of building AI solutions. You can now apply edits to multiple inputs in New Portal simultaneously. Simply select the inputs you would like to edit, and make your changes, and we will apply the edits to all selected inputs.
Quickly sort and analyze your training data. You can now filter your inputs based on the dataset that they came from, and this gives you powerful insight and control over the data you use to train your models.