Graphsignal – Machine learning profiler for training and inference
Hi HN, I'm the founder of Graphsignal (https://graphsignal.com). Graphsignal is a machine learning profiler. We've created it to make ML profiling simple and usable. It provides performance summaries, ML operation and kernel level statistics as well as detailed resource usage information necessary for making training and inference faster and more efficient. Profilers help fix performance issues, improve user experience and reduce computation costs. Such improvements benefit machine learning profoundly; model training jobs that run for hours or days could be made much shorter and…
In plain words
Graphsignal is a machine learning profiler that helps developers optimize model training and inference. It provides performance summaries, operation and kernel-level statistics, and detailed resource usage data to identify bottlenecks and improve speed and efficiency. The tool is designed for machine learning engineers and data scientists working with training jobs or inference systems who need to reduce computation costs and latency.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hi HN, I'm the founder of Graphsignal (https://graphsignal.com). Graphsignal is a machine learning profiler. We've created it to make ML profiling simple and usable. It provides performance summaries, ML operation and kernel level statistics as well as detailed resource usage information necessary for making training and inference faster and more efficient. Profilers help fix performance issues, improve user experience and reduce computation costs. Such improvements benefit machine learning profoundly; model training jobs that run for hours or days could be made much shorter and inference latency could be reduced resulting in significantly lower costs and improved user experience. I realized the benefits in one of my previous projects, where the model would have to be trained regularly and be used for inference on huge amount of data. Having spent last decade developing profiling and monitoring tools, it seemed logical for me to use a profiler for the task. But since the training and inference were running remotely, I had a hard time using existing ML profilers. TensorFlow and PyTorch provide built-in ML profilers, which utilize NVIDIA's profiling interface (CUPTI) under the hood for GPU profiling. One way to use those profilers is via locally installed TensorBoard or by logging the profiles. In turn, Graphsignal Profiler (https://github.com/graphsignal/graphsignal) uses the built-in profilers as well as other tools to enable automatic profiling in any environment, including notebooks, training pipelines, periodic batch jobs, model serving and so on, without installing additional servers/software. It also allows teams to share and collaborate online. Basically, the profiles along with environment and usage information are be automatically recorded and sent to Graphsignal where they are available for analysis. Trying it out is easy: 1) sign up for a free account; 2) add the profiler to your ML code and run it; 3) see and analyze the profiles at graphsignal.com. Everything is described in the Quick Start Guide https://graphsignal.com/docs/profiler/quick-start/. I'm very excited to show it to you here and will appreciate any thoughts, comments and feedback!
Does a similar job
all alternatives →- GMGraphsignal – ML profiler to speed up training and inference2022 · github.com · ▲16
Hi, Graphsignal founder here. We've launched Graphsignal earlier this year to make machine learning profiling practical and easy to use. Basically, it enables the profile-optimize-benchmark loop. For example, making inference faster by optimizing an ML model, while still maintaining accuracy. We've make a lot of progress that I wanted to share. The profiler now natively supports TensorFlow, Keras, PyTorch, PyTorch Lightning, Hugging Face, XGBoost and JAX frameworks along with built-in support for distributed workloads. Profiles now include tracing information in chrome trace format. Process…
- GPGraphsignal – Production Model Monitoring: Anomalies, Drift, Exceptions2021 · github.com · ▲30
- GRgraphql-request-profiler – a small GraphQL performance analysis plugin2022 · github.com · ▲17
- FCFully client-side GPT2 prediction visualizer2023 · perplexity.vercel.app · ▲153
Hi HN! I've found this visualization tool immensely helpful over the years for getting an intuition for how an LLM "sees" some piece of text, and with a bit of elbow grease decided to move all compute to client side so I could make it publicly available. I've found it particularly useful for - Understanding exactly how repetition and patterns affect a small LM's ability to predict correctly - Understanding different tokenization patterns and how it affects model output - Getting a general sense of how "hard" different prediction tasks are for GPT-style models Known problems (that I probably…
- DFData Formulator – AI-powered data visualization from Microsoft Research2024 · github.com · ▲212
Creating data visualizations with AI nowadays often means chat, chat and more chats...and writing long prompts can be annoying while they are also not the most effective way to describe your visualization designs. Data Formulator blends UI interaction with natural language so that you can create visualizations with AI much more effectively! You can: * create rich visualizations beyond initial datasets, where AI helps transforming and visualizing data along the way * iterate your designs and dive deeper using data threads, a new way to manage your conversation with AI. Here is a demo video:…
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 19d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, March 2022
the whole month →- IM
Hi HN! I made many small web games in the past with Unity or Phaser. But this time I wanted to make something completely on my own. So I coded a basic game framework in plain JavaScript, and used it to make Almost Pong. This was a really fun project that taught me a lot about JavaScript, and I plan to make more minimalist games with this framework. Interesting fact: Almost Pong doesn't load any assets, all sprites and sounds are generated with code. Happy to answer questions, and please let me know if you have any feedback on the game. Thanks!
Dev tools · 2022 · lessmilk.com
- IM
Life & fun · 2022 · networksfromscratch.com
- HA
Paste the following into the console of any HN page - for annotated avatars on all HN comments. (self contained code) for(u of document.querySelectorAll('.hnuser'))for(u.prepend(c=document.createElement('canvas')),x=c.getContext('2d'),c.width=18,c.height=14,s=u.innerText,r=1,i=28+s.length;i--;i>>29>X*X/3+Y/2&&x.fillRect(6+2*X,2*Y,2,2)&x.fillRect(6-2*X,2*Y,2,2):r+=s.charCodeAt(i-28,x.fillStyle='#'+(r>>8&0xFFFFFF).toString(16)))r^=r>>17,r^=r>2
Dev tools · 2022
- AG
Life & fun · 2022 · backofyourhand.com