whylogs
graphsignal-python
whylogs | graphsignal-python | |
---|---|---|
6 | 30 | |
2,548 | 200 | |
0.9% | 1.0% | |
9.0 | 8.1 | |
3 days ago | 6 days ago | |
Jupyter Notebook | Python | |
Apache License 2.0 | Apache License 2.0 |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
whylogs
-
The hand-picked selection of the best Python libraries and tools of 2022
whylogs — model monitoring
-
Data Validation tools
Have a look at whylogs. Nice profiling functionality incl. definition of constraints on profiles: https://github.com/whylabs/whylogs
- [D] Open Source ML Organisations to contribute to?
- whylogs: The open standard for data logging
-
I am Alessya Visnjic, co-founder and CEO of WhyLabs. I am here to talk about MLOps, AI Observability and our recent product announcements. Ask me anything!
WhyLabs has an open-source first approach. We maintain an open standard for data and ML logging https://github.com/whylabs/whylogs, which allows anybody to begin logging statistical properties of data in their data pipeline, ML inference, feature stores, etc. These statistical profiles capture all the key signals to enable observability in a given component. This unique approach means that we can run a fully SaaS service, which allows for huge scalability (in both the size of models and their number), and ensures that our customers are able to maintain their data autonomy. We maintain a huge array of integrations for whylogs, including Python, Spark, Kafka, Ray, Flask, MLflow, Kubeflow, etc… Once the profiles are captured systematically, they are centralized in the WhyLabs platform, where we organize them, run forecasting and anomaly detection on each metric, and surface alerts to users. The platform itself has a zero-config design philosophy, meaning all monitoring configurations can be set up using smart baselines and require no manual configuration. The TL;DR here is the focus on open source integrations, working with data at massive/streaming scale, and removing manual effort from maintaining configuration.
-
Machine learning’s crumbling foundations – by Cory Doctorow
This is why we've been trying to encourage people to think about lightweight data logging as a mitigation for data quality problems. Similar to how we monitor applications with Prometheus, we should approach ML monitoring with the same rigor.
Disclaimer: I'm one of the authors. We spend a lot of effort to build the standard for data logging here: https://github.com/whylabs/whylogs. It's meant to be a lightweight and open standard for collecting statistical signatures of your data without having to run SQL/expensive analysis.
graphsignal-python
-
Show HN: Python Monitoring for LLMs, OpenAI, Inference, GPUs
We've built it for apps that use LLMs and other ML models. The lightweight Python agent autoinstruments OpenAI, LangChain, Banana, and other APIs and frameworks. Basically by adding one line of code you'll be able to monitor and analyze latency, errors, compute and costs. Profiling using CProfile, PyTorch Kineto or Yappi can be enabled if code-level statistics are necessary.
Here is a short demo screencast for a LangChain/OpenAI app: https://www.loom.com/share/17ba8aff32b74d74b7ba7f5357ed9250
In terms of data privacy, we only send metadata and statistics to https://graphsignal.com. So no raw data, such as prompts or images leave your app.
-
Show HN: Python Monitoring for AI: LLMs, OpenAI, Inference, GPUs
Hi HN. I'm excited to share our AI-focused application monitoring and analytics for Python!
We've built it for apps that use LLMs and other ML models. The lightweight Python agent autoinstruments OpenAI, LangChain, Banana, and other APIs and frameworks. Basically by adding one line of code you'll be able to monitor and analyze latency, errors, compute and costs. Profiling using CProfile, PyTorch Kineto or Yappi can be enabled if code-level statistics are necessary.
Here is a short demo screencast for a LangChain/OpenAI app: https://www.loom.com/share/17ba8aff32b74d74b7ba7f5357ed9250
In terms of data privacy, we only send metadata and statistics to https://graphsignal.com. So no raw data, such as prompts or images leave your app.
We'd love to hear your feedback or ideas!
-
[N] Monitor OpenAI API Latency, Tokens, Rate Limits, and More with Graphsignal
Here is a blog post with more info and screenshots: Monitor OpenAI API Latency, Tokens, Rate Limits, and More. And the GitHub repo.
-
Monitor OpenAI API Latency, Tokens, Rate Limits, and More
Relying on hosted inference with LLMs, such as via OpenAI API, in production has some challenges. The use of APIs should be designed around unstable latency, rate limits, token counts, costs, etc. To make it observable we've built tracing and monitoring specifically for AI apps. For example, the OpenAI Python library is monitored automatically, no need to do anything. We'll be adding support for more libraries. If you'd like to give it try, see https://github.com/graphsignal/graphsignal or the docs.
-
[N] Easily profile FastAPI model serving
We've added a simple way to profile any model serving endpoint, including FastAPI, to identify bottlenecks and make inference (incl. data processing) faster, especially for big models and data. Wanted to share it here in case someone is struggling with profiling and monitoring of deployed code and models. By default, generic Python profiler will automatically profile some of the inferences (and measure all inferences). You can also specify other profilers for PyTorch, TensorFlow, Jax and ONNX Runtime. All profiles and metrics will be available on the SaaS dashboard, no need to setup anything. A couple of links to get started: Repo: https://github.com/graphsignal/graphsignal FastAPI example: https://graphsignal.com/docs/integrations/fastapi/ Happy for any feedback!
-
[P] Using Sparsity & Clustering to compress your models: Efficient Deep Learning Book
Thanks for sharing! That's a very timely topic. I've actually created a profiler to track and analyze inference optimizations, i.e. enable the optimize-verify-evaluate loop.
-
[N] Accuracy-Aware Inference Optimization Tracking and Profiling
To address all these problems, we've built a tool to track inference optimizations, see how accuracy is affected, verify that the optimizations were applied and locate any bottlenecks for further improvements. All in one place.
- Show HN: Graphsignal – ML profiler to speed up training and inference
What are some alternatives?
evidently - Evaluate and monitor ML models from validation to production. Join our Discord: https://discord.com/invite/xZjKRaNp8b
metaflow - :rocket: Build and manage real-life ML, AI, and data science projects with ease!
seldon-core - An MLOps framework to package, deploy, monitor and manage thousands of production machine learning models
second-brain-agent - 🧠 Second Brain AI agent
flyte - Scalable and flexible workflow orchestration platform that seamlessly unifies data, ML and analytics stacks.
Imaginer - Imagine with AI
datatap-python - Focus on Algorithm Design, Not on Data Wrangling
Keras - Deep Learning for humans
langchain - ⚡ Building applications with LLMs through composability ⚡ [Moved to: https://github.com/langchain-ai/langchain]
tensor-sensor - The goal of this library is to generate more helpful exception messages for matrix algebra expressions for numpy, pytorch, jax, tensorflow, keras, fastai.
Activeloop Hub - Data Lake for Deep Learning. Build, manage, query, version, & visualize datasets. Stream data real-time to PyTorch/TensorFlow. https://activeloop.ai [Moved to: https://github.com/activeloopai/deeplake]
PhoneTracer - Gets GPS location of phone numbers