Small model fine-tuning
‍made simple

Train a task-specific natural language processing model using just a few dozen examples
Schedule a demo

NLP is our game

Core NLP use cases: classification, extractive question-answering, NER, summarization, ranking.

Low-data input

We only required a few dozen annotated datapoints to train a performant model.

Fully automated

We automate the processes of
fine-tuning and benchmarking your model.

On-prem or via API

You can host the model yourself or simply access it via our API.

How does it work?

We use knowledge distillation to train task-specific, small language models for your particular problem
Learn more
1
Upload your data
We need a “task description”, a few dozen examples of the task the model is supposed to perform, and any additional data about the problem (documentation, unlabelled examples, …)
2
Evaluate the model
We train the model and share benchmarks based on a test dataset we isolate from the training data. You can either iterate (change the task description, upload more data) or choose to Deploy the model
3
Access and integrate the model
We can either share the model binaries with you or handle the deployment and provide an API endpoint that you can query & integrate with your product.

Key benefits

Best of both worlds – LLMs & conventional Machine Learning

Accelerating your development process

You need a much smaller labeled training dataset than ever before; there is no need to onboard, manage and pay for subject matter expert human annotators. Few dozen labeled examples instead of 10 000s.

Reducing latency & costs of your AI pipeline

Using a smaller specialized model enables deployment on cheaper and faster infrastructure. Models fine-tuned to your use case don't need as much context in the prompt, further reducing your token usage.

Local deployment

Small models can be easily deployed directly on mobile hardware as a part of an application. This means your applications are not reliant on a strong network connection and it helps ensure data privacy compliance.
100
x
Less data needed to train a performant model
Same accuracy using significantly less data thanks to model distillation
50
x
Smaller model with the same accuracy
Specialized Llama3 8b compared to instruction tuned Llama3 405b
10
x
Faster model inference
Specialized Llama3 8b compared to instruction tuned Llama3 405b
12
h
Needed to train and deploy specialized model
100x Less time compared to a standard 7-week annotation & training project

Let’s build a model for your business

Create a custom Small Language Model for your AI product
Schedule a demo
Contact us
© disitil labs 2024
Home
How it works
About us
FAQ
Contact