This product was not featured by Product Hunt yet. It will not be visible on their landing page and won't be ranked (cannot win product of the day regardless of upvotes).
Product upvotes vs the next 3
Waiting for data. Loading
Product comments vs the next 3
Waiting for data. Loading
Product upvote speed vs the next 3
Waiting for data. Loading
Product upvotes and comments
Waiting for data. Loading
Product vs the next 3
Loading
Gradian
Find the data that broke your fine-tune.
Training-data attribution for LoRA fine-tuned LLMs. Gradian scores every training example against the capability you lost, then runs a battery of dataset and config checks for the failures influence math misses.
Over the past few months, I've been spending a lot of time fine tuning LLMs, and I kept hitting the same frustrating problem. The model would get worse, but I had no idea why.
Nothing crashed. The loss curve looked normal. Training finished successfully. But suddenly the model would stop mid sentence, lose capabilities it had before, or start giving confidently wrong answers.
Debugging meant changing one thing at a time, retraining, and then waiting hours to see if I had guessed correctly. Most of the time I hadn't.
So I started building the tool I wished I had.
The first question I wanted to answer was simple. Which training examples actually caused a capability to improve or get worse?
That led me into gradient based data attribution. Along the way I realized many of the biggest issues had nothing to do with the data itself. They came from subtle training mistakes like truncated completions, missing EOS tokens, loss being computed on prompts, train and eval contamination, or hyperparameters copied from a completely different training recipe.
The biggest surprise came while building the attribution method itself. My first version consistently ranked intentionally poisoned examples as some of the most helpful examples in the dataset. It turned out they were teaching the model the output format perfectly, even while teaching the wrong facts. That forced me to rethink the approach, and eventually led to subtracting the gradient of the model's actual prediction. The rankings immediately started matching reality.
Gradian is the result. It's an open source tool that helps explain why a fine tune changed, instead of leaving you guessing.
I'd love to hear your stories. If you've ever had a fine tune mysteriously get worse, what ended up being the cause?
Thanks so much for checking out Gradian! 🚀
About Gradian on Product Hunt
“Find the data that broke your fine-tune.”
Gradian was submitted on Product Hunt and earned 3 upvotes and 1 comments, placing #93 on the daily leaderboard. Training-data attribution for LoRA fine-tuned LLMs. Gradian scores every training example against the capability you lost, then runs a battery of dataset and config checks for the failures influence math misses.
On the analytics side, Gradian competes within Developer Tools, Artificial Intelligence, GitHub and Data Science — topics that collectively have 1M followers on Product Hunt. The dashboard above tracks how Gradian performed against the three products that launched closest to it on the same day.
Who hunted Gradian?
Gradian was hunted by Abdullah Abdelrazek. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
For a complete overview of Gradian including community comment highlights and product details, visit the product overview.
Abdullah Abdelrazek
Hey Product Hunt! 👋
I'm Abdullah, the creator of Gradian.
Over the past few months, I've been spending a lot of time fine tuning LLMs, and I kept hitting the same frustrating problem. The model would get worse, but I had no idea why.
Nothing crashed. The loss curve looked normal. Training finished successfully. But suddenly the model would stop mid sentence, lose capabilities it had before, or start giving confidently wrong answers.
Debugging meant changing one thing at a time, retraining, and then waiting hours to see if I had guessed correctly. Most of the time I hadn't.
So I started building the tool I wished I had.
The first question I wanted to answer was simple. Which training examples actually caused a capability to improve or get worse?
That led me into gradient based data attribution. Along the way I realized many of the biggest issues had nothing to do with the data itself. They came from subtle training mistakes like truncated completions, missing EOS tokens, loss being computed on prompts, train and eval contamination, or hyperparameters copied from a completely different training recipe.
The biggest surprise came while building the attribution method itself. My first version consistently ranked intentionally poisoned examples as some of the most helpful examples in the dataset. It turned out they were teaching the model the output format perfectly, even while teaching the wrong facts. That forced me to rethink the approach, and eventually led to subtracting the gradient of the model's actual prediction. The rankings immediately started matching reality.
Gradian is the result. It's an open source tool that helps explain why a fine tune changed, instead of leaving you guessing.
I'd love to hear your stories. If you've ever had a fine tune mysteriously get worse, what ended up being the cause?
Thanks so much for checking out Gradian! 🚀