Built by AI leaders from Uber, Google, Apple and Amazon. Developed and deployed with the world’s leading organizations.
Bigger Isn’t Always Better
Fine-tune smaller task-specific LLMs that outperform bloated alternatives from commercial vendors. Don’t pay for what you don’t need.
The Fastest Way to Build & Deploy Custom Models
Fine-tune and serve any open-source LLM —all within your environment, using your data, on top of a proven scalable infrastructure. Built by the team that did it all at Uber.
Privately Deploy Any Open Source LLM
Stop spending hours wrestling with complex model deployments. Deploy and query the latest open-source pre-trained LLM—like Llama-2, Mistral and Falcon—on scalable managed infrastructure in your VPC or Predibase cloud. All of this can be achieved in minutes with just a few lines of code.
# Deploy an LLM from Huggingface llm = pc.LLM("hf://meta-llama/Llama-2-13b-hf") llm.deploy(deployment_name="llama-2-13b") # Prompt the deployed LLM deployment = pc.LLM("pb://deployments/llama-2-13b") deployment.prompt( "Write an algorithm in Java to reverse the words in a string." )
Efficiently Fine-tune Models for Your Task
No more out-of-memory errors or costly training jobs. Fine-tune any open-source LLM on the most readily available GPUs using Predibase’s optimized training system. We automatically apply optimizations such as quantization, low-rank adaptation, and memory-efficient distributed training combined with right-sized compute to ensure your jobs are successfully trained as efficiently as possible.
# Specify a Huggingface LLM to fine-tune llm = pc.LLM("hf://meta-llama/Llama-2-13b-hf") # Kick off the fine-tune job job = llm.finetune( prompt_template=prompt_template, target="output", dataset="s3_bucket/code_alpaca", repo="finetune-code-alpaca" ) # Stream training logs and metrics job.get()
Dynamically Serve Many Fine-tuned LLMs In One Deployment
Our scalable serving infra automatically scales up and down to meet the demands of your production environment. Dynamically serve many fine-tuned LLMs together for over 100x cost reduction versus dedicated deployments with our novel LoRA Exchange (LoRAX) architecture. Load and query them in seconds.
Read more about LoRAX.
# Specify a base model and fine-tuned model base_model = pc.LLM("pb://deployments/llama-2-13b") finetuned_model = pc.get_model("finetune-code-alpaca") # Prompt the fine-tuned model instantly using LoRAX finetuned_deployment = base_model.with_adapter( finetuned_model ) finetuned_deployment.prompt( "Write an algorithm in Java to reverse the words in a string." )
Production AI in Record Time
Customize and ship models faster with the first end-to-end AI platform that’s designed for engineers.
Your Models, Your Property
Start owning and stop renting ML models. The models you build and customize on Predibase are your property, deployed securely inside your VPC, with full data privacy.
Designed for Developers
Built by developers, for developers. Predibase enables any software engineer to do the work of an ML engineer in an easy-to-understand declarative way.
Managed Serverless Infrastructure
Stop wasting time managing distributed clusters–get fully managed, optimized compute configured for your needs without all the time and cost.
Built on Proven Open Source Technology
Ludwig is a deep learning toolbox to declaratively develop, train, fine-tune, test and deploy state-of-the-art models. Ludwig puts deep learning in the hands of analysts, engineers and data scientists without requiring low-level ML code.
Horovod is a distributed deep learning framework that scales PyTorch and TensorFlow training to hundreds of machines. It supports Tensorflow, Keras, Pytorch, Apache MXNet and has been used to productionize deep learning models across industries
Predibase works across any supervised machine learning use case, so if you have labeled or historical data, our platform can learn from those patterns and apply it to use cases such as: