π¬ Free, local-first experiment tracking for ML researchers π€ Drop-in replacement for wandb | Built by @huggingface π Track, visualize & share your training
Joined August 2025
- Tweets86
- Following16
- Followers302
- Likes87
TrackioApp retweeted
Comparing multiple models has never been this easy, thanks to gr.Workflowπ
We used ML Intern to distill FLUX.3 Image Klein's 9B prompt upsampler into 0.8B and 2B models. Short prompt in β full prompt + bounding boxes out. The 0.8B can even run on a CPU.
Links in reply.
TrackioApp retweeted
i gave astra a robot, a paint brush, and a camera then asked it to paint the golden gate bridge in real life!
it figured out how to control the robot, and progressively got better throughout its attempts. the timelapse is sick
TrackioApp retweeted
From there, ML Intern can create repos, generate datasets, train models and monitor progress as jobs run.
Every training run gets a @TrackioApp dashboard so you can follow along.
TrackioApp retweeted
We just released trackio 0.36, which supports logging a new data type! Can you tell what we added?
TrackioApp retweeted
We are super excited to release Trackio Logbooks: the perfect way to capture all of your AI experiments as an optimized static html bundle π₯
Upgrade to Trackio 0.34.0 and start recording your code, agent traces, and artifacts automatically (and reproducibly)!
Docs: github.com/gradio-app/trackiβ¦
TrackioApp retweeted
I'm very excited to share that @nvidia just released Nemotron-3-Embed: two multilingual embedding models for retrieval
The 8B takes the top spot on RTEB.
- Nemotron-3-Embed-1B-BF16 (2048-dim)
- Nemotron-3-Embed-8B-BF16 (4096-dim)
Both OpenMDW-1.1, ready for commercial use.
π§΅
TrackioApp retweeted
Join the ICML reproduction 2026 challenge leaderboard for a chance to win up to $4k in GPU credits and prizes:
huggingface.co/spaces/ICML-2β¦
TrackioApp retweeted
70% of AI papers arenβt reproducible. With ICML 2026 happening last week, a lot of incredible research is dropping, but how much of it is actually real and reproducible? With @askalphaxiv, we are hosting a community challenge to find out.
Pick an ICML paper, deploy an agent, and help build a massive library of open artifacts.
Up to $4k in GPU credits for the best reproductions!
Learn more here: huggingface.co/spaces/ICML-2β¦
TrackioApp retweeted
TruffleHog now supports scanning secrets in @huggingface Storage Buckets.
As AI agents increasingly crawl code, datasets, and storage, preventing leaked credentials is becoming more important than ever.
Great collab with the @trufflesecurity team! π
TrackioApp retweeted
Lots of reproductions happening already! Join the effort (and bring your agents): huggingface.co/spaces/ICML-2β¦
TrackioApp retweeted
π Trackio Artifacts is officially LIVE! π¦π₯
The #1 most requested feature by the Trackio community is finally here: Full support for logging versioned Artifacts (datasets, models, arbitrary files, or binary blobs).
If youβve been looking for a seamless, lightweight way to manage your ML assets without the bloat, this is it. Best of all? We designed it with a zero-friction migration path. Itβs the exact same API signature you already know and love from wandb, making it a literal drop-in replacement:
TrackioApp retweeted
We're open-sourcing a new benchmark on @huggingface: IFStruct
The goal is to measure output validity and schema following across diverse prompts.
We also detail how we generated it and the design choices behind it in our blog post.
TrackioApp retweeted
Use Trackio.
Happy to release the latest version of @TrackioApp which includes Brotli compression and other nice features which makes your dashboard a lot snappier!
I recommend upgrading, especially if you have a lot of runs!
pip install --upgrade trackio
TrackioApp retweeted
I didn't realize this was a feature - great idea
A study from @Stanford showed that 71.3% of chatgpt queries could be accurately answered by a local model. I suspect a major part of enterprise AI workloads could be run locally too for free (compared to the massive costs of frontier API cost).
Also, it reduces the risk of these workloads being taken away from you because you own the models instead of renting them - which sounds like a good idea these days haha.
That's why we're introducing the ability for everyone to filter AI models on @huggingface based on your local hardware.
For me, there are 800k+ public models that fit on my M5 24GB and that I can use easily thanks to llamacpp.
Let's go local AI!
TrackioApp retweeted
In recent months, @Gradio has about 1 million active developers, but we've decided to build something new.
A new way to build ML apps that is intuitive for both humans and agents. A new workflow that could get a million daily users on its own.
Stay tuned.
TrackioApp retweeted
We just released Gradio 6.15 with a big change: a Node-based proxy server to Gradio apps
When server-side rendering is enabled, we have a smart router which reduces the load on the main FastAPI server significantly, which means that Gradio apps with many concurrent users will now feel much faster.
No code changes needed. Enjoy!
TrackioApp retweeted
Our primary goal with @TrackioApp right now is data integrity at very, very high throughputs.
So if you (or your agents) are running many ML experiments, you can be sure your logs are saved whether that's local or pushed to Spaces!
TrackioApp retweeted
We just released @TrackioApp 0.17, with a major focus on throughput!
In my tests, you should be able to 1000 metrics every ~4 seconds when pushing to Spaces, and basically unlimited throughput when logging locally.
ππ πππ πππππππ --πππππππ πππππππ
