Stochastic X: Community Version

Nov 18, 2022

Accelerating Access to the Best AI

At Stochastic, our vision is to make the best AI accessible to everyone.

We plan to achieve this vision through Stochastic X, our scalable deep learning platform that companies can use to easily build, accelerate, and deploy the toughest deep learning models for the best performance, at a fraction of cloud costs.

AI models like GPT-3 (large language model) and DALL-E (text-to-image model) are powerful but difficult to develop and deploy into production. We leverage State-Of-The-Art techniques to automatically compress and accelerate these models, compiling efficient cloud systems to help companies reduce the time-to-market and achieve the best performance while minimizing cloud costs without highly skilled engineers. For many, accelerated speed could enable new product possibilities and give them the competitive advantage they need.

Introducing the Community Version

We are pleased to introduce the Community Version of Stochastic X! The Community Version provides Stochastic’s key platform features for free to the community. With the Community Version, users can easily install Stochastic's software on their computer to run the platform locally on their device.

The main features in the Community Version are model training, benchmarking, deployment, and inferencing

  • Model training: Readily fine-tune and train models using your own datasets. Fine-tuning jobs can be independently deployed so they work as an online learning job. Enterprise users could utilize this feature to service their own customers
  • Benchmarking: Easily compare model performance against different compute options and datasets
  • Deployment: Quickly deploy your models into production with support for CPU and GPU
  • Inference: Experience reduced inference times with your models

To access the Community Version of Stochastic X, sign up here!

Model Training

Using the Community Version, users can fine-tune their machine learning models with their own datasets. To fine-tune models using the Community Version, users just have to select a dataset and model, and then Stochastic's software will automatically fine-tune the model for them.

stochasticx finetuning launch sequence_classification \
  --name finetuning_job_name \
  --model_name_or_id distilbert-base-uncased \
  --dataset_name_or_id mrpc \
  --sentence1_column "sentence1" \
  --sentence2_column "sentence2" \
  --label_column "label"

Benchmarking

The Community Version's automatic benchmarking tool allows users to compare model performance against local dataset and compute options. With the Community Version, users are only allowed to benchmark compute options on their local machine. To benchmark compute options outside of their local machines, users must sign up for early access to Stochastic Cloud.

stochasticx benchmarking start

Deployment

With Stochastic's Community Version, users can deploy their fine-tuned models locally, with a few lines of code.

stochasticx deployments deploy sequence_classification \
  --model_name_or_id finetuning_job_name_optimized \
  --model_type hf

Inference

Stochastic's Community Version allows users to access accelerated inferencing. With this feature, users can experience accelerated model inferencing to experience reduced inference times.

stochasticx inference sequence_classification \
  --deployment_id d5063658-ac67-42cd-bbdb-7ec653b79c94 \
  --text "I love Stochastic"


Cloud and Enterprise

Stochastic Cloud exists for users who want to offload any local tasks to the cloud, with many advanced features only available on the Enterprise version. To access the Stochastic Cloud, users must first request early access.

Stochastic Cloud

Stochastic Cloud allows users to access the Stochastic AI platform on the cloud. The features accessible to Stochastic Cloud users are the Web UI, accelerated fine-tuning, advanced benchmarking, automated model compression, and auto-scaling.

  • Web UI: The main feature of both Stochastic Cloud and Enterprise is the Web UI, where users can deploy and fine-tune models through an interactive interface
  • Accelerated Fine-tuning: reduce the amount of training time required for models
  • Advanced Benchmarking: compare models against local compute options and cloud compute options across both GPU and CPU
  • Automated Model Compression: automatically reduce compute costs during inference, reducing memory demand
  • Auto-scaling: automatically adjust the needed resources for each pipeline, minimizing cost

Enterprise

Stochastic's Enterprise version supports enterprises that want to host Stochastic's software locally and already have existing ML infrastructure or plan to develop one. The Enterprise Version enables all of the features available on Stochastic Cloud, with additional premium features.

To learn more about enterprise solutions, reach out: Enterprise Inquiries →

Conclusion

Stochastic X is the next-generation deep learning platform built for everyone to access the best AI. The launch of the Community Version will allow users to access Stochastic’s core features for free!

Access our product here: Sign Up →

Contact our team: Contact →