Taking deep learning to production with RedisAI
summary
RedisAI is a Redis module that adds tensors & graphs as Redis data types and executes models on CPU and GPU using PyTorch TensorFlow ONNXRuntime backends while exposing a tensor API and TorchScript-like scripting. The demo shows a model binary loaded with modelset, tensor I/O and modelrun executing in a highly available Redis setup with replication and clustering, plus automated batching sharding and streams on the roadmap toward 1.0 and working sample code and a live deployment demo.
AI-written summary · about
chapters (34)
- 0:00Introduction
- 2:03What is Software 20
- 2:40Why we decided to take this up
- 3:19Our plan
- 3:36Hangar
- 4:40Python package
- 5:28What do we already have
- 5:43Deep learning 101
- 6:30Deep learning frameworks
- 7:09What next
- 7:24Existing production strategies
- 8:37Production requirements
- 10:07What is Redis
- 11:59Redis mode
- 15:48Architecture
- 16:40API Soph ready
- 18:04Tensor structure
- 19:07Model set
- 19:52Model run
- 20:21Script
- 23:21Replication
- 24:03Simple example
- 25:20Redis client libraries
- 25:57Advantages
- 28:49RedisAI repository
- 29:06DAG
- 29:40Volatile
- 30:22Readonly
- 31:43Blocking line
- 32:15Streams
- 32:34Automatic
- 33:06ReadyAI Enterprise
- 33:18ReadyAI Blog
- 33:34Outro
related talks
24mWhen Real Time Meets Big Data & Analytics
59mUsing Redis beyond caching
19mThe Key Role Of Feature Stores In Deploying Real-time AI
43mThe Challenges of Deploying Real-time AI for Finance and how Open Source can help
14m5 great ways to reimagine your data layer
19mML feature store a cornerstone in modern data architecture for AI/ML
43mDeep learning beyond the learning
26mEngineering Excellence: Platforms and Architectures for Real-Time AI Analytics