Example Notebooks
These example notebooks demonstrate how to use Cobalt across a variety of real-world scenarios. Each notebook is available in the cobalt-examples GitHub repository.
E-Commerce Embedding Model Comparison
Apply Cobalt’s model comparison table to choose the best embedding model for your e-commerce vector database. Cobalt intelligently clusters together users’ product search queries into valuable and interpretable categories on which to compare different models.
View on GitHub | View in Google Colab
E-Commerce Embedding Model Comparison (Fine-Tuning)
Similar to the above, but focused on tradeoffs from fine-tuning. The two models are a base E5 model from Hugging Face and its fine-tuned version using Marqo’s Marqtune platform. Cobalt reveals performance tradeoffs from fine-tuning through the model comparison table.
View on GitHub | View in Google Colab
Basic Tutorial
A simple example using a synthetic tabular dataset and a scikit-learn-based model to illustrate the main parts of the Cobalt interface. The same example is outlined in the Tutorial.
Text Classification
Uses Cobalt to explore and debug a transformer-based text classification model from Hugging Face. Requires the transformers package to be installed.
Data Drift
A simple illustration of how Cobalt can help diagnose and understand data drift even in the absence of a model or model performance metrics.
Image Clustering with CLIP
Demonstrates the use of Cobalt to make sense of the ImageNette dataset and explore its target labels using embeddings generated by the CLIP model.
This notebook requires the OpenAI clip package to be installed, which includes torch, torchvision, etc.