Uptrain

We're excited to announce the super-fast release of UpTrain version 0.2, packed with exciting new features and enhancements to supercharge your ML model monitoring experience! Here's what's new:

✨ Automated Fine-tuning with WeightWatcher: Introducing automated fine-tuning using the WeightWatcher package (https://github.com/CalculatedContent/WeightWatcher). This allows for effortlessly optimizing ML models for improved performance and accuracy by observing the weights of your pre-trained model.

🔪 Feature Slicing: Evaluate the performance of your models on different categories of features with the new feature slicing capability.

📊 New Logging API: We've introduced a powerful new logging API to streamline the ML model monitoring process.

💼 Enterprise Edition (ee) Folder: For enterprise users, we've introduced the ee folder, providing enhanced efficiency and scalability for managing large-scale ML monitoring operations. This allows enterprises to monitor their models at scale with ease.

🌟 Golden Testing Dataset: Now, you can evaluate new LLM models and prompts with confidence using a curated golden testing dataset. This ensures that freshly minted models meet the highest standards of performance and accuracy.

Upgrade the UpTrain package now and take your ML model monitoring to the next level! 🚀🔥

Note: As always, we value your feedback and are committed to delivering the best ML monitoring experience. Please share your thoughts and suggestions with us to help us continue improving UpTrain. Happy monitoring! 😊👩‍💻👨‍💻

Minor production bug fixes related to data integrity.

🚀 UpTrain Feature Release 🚀

We're excited to announce a new release of UpTrain with some exciting features! Here's what's new:

📝 Grammar Check: We've added a new integrity check to our Language Model Monitoring (LLM) outputs. Now, you can check for grammatical errors in your text data to ensure it meets your quality standards.

🔗 Feature Concatenation: We've introduced a new feature concatenation measurable that allows adding monitors on a subset of columns. This will help you measure and monitor the performance of specific subsets of features.

🎤 Automatic Speech Recognition: We've added an example for automatic speech recognition (ASR). With this new feature, you can now monitor the performance of your ASR models and catch issues early.

📈 Hamming Distance: We've added Hamming distance as a distance metric. Now you can use this metric to measure the similarity between binary or categorical data.

We hope these new features help you monitor your ML models more effectively! As always, we welcome your feedback and suggestions for future releases.

We are excited to announce the latest release of UpTrain with a range of new features to help you better monitor and maintain your machine learning models. Here are some highlights of the new features:

🔍 HDBSCAN Clustering: With our new HDBSCAN clustering option, you can now perform hierarchical density-based clustering on your data with ease.

📈 Z-Score Data Integrity: We've added z-score data integrity to help you identify outliers using statistical methods, making it easier to maintain data integrity.

🔄 Feature Drift with PSI: Our new feature drift using PSI allows you to measure the drift between your training and production data, helping you detect and address potential issues early on.

🆔 Data-point Identification: We've added a new feature to easily identify data-points by their IDs in our data visualizations.

💬 Conversation Summarization Example: We've added a new example showcasing how to use UpTrain to summarize conversations, providing a practical use case for NLP applications.

🧪 Tests for Concept Drift Algorithms: We've also added tests for the ADWIN and DDM concept drift algorithms, ensuring that your models remain accurate and reliable over time.

Upgrade to the latest version of UpTrain today to take advantage of these new features and streamline your machine learning monitoring and maintenance. As always, please let us know if you have any questions or feedback. Happy training!

🚀 New UpTrain Release 🚀

In the next feature release of UpTrain, users can look forward to several exciting updates and improvements that enhance the software's functionality 🚀. This new release focuses on enhancing user experience, improving performance, and expanding the software's capabilities to better serve the needs of the machine learning community 🌟.

- Added documentation to the GitHub repository 📚: Makes it easier for users and contributors to understand and improve UpTrain by providing comprehensive documentation on the GitHub repository.
- Removed duplicated code from t-SNE and UMAP 🧹: Streamlines the codebase by eliminating redundant code, leading to a more efficient and maintainable software package.
- Logging and monitoring in background process ⏳: Enables users to log and monitor model performance and other metrics without interrupting their workflow.
- Added concept drift detection with ADWIN 📊: Enhances concept drift detection by incorporating the ADaptive WINdowing (ADWIN) algorithm for more accurate and responsive drift detection.
- Removed package dependency on UMAP and SHAP 🛠️: Simplifies the installation process and reduces potential conflicts by removing the need for the UMAP and SHAP packages.
- Better naming for edge cases 🏷️: Introduces clearer naming conventions for edge cases, making identifying and managing them easier.
- Logging Args put together 🗂️: Streamlines argument logging, making it easier to manage and track settings for your machine learning models.
- Several bug-fixes 🐛: Addresses various issues to improve the overall stability and performance of UpTrain.