LogoLogo
  • MLServer
  • Getting Started
  • User Guide
    • Content Types (and Codecs)
    • OpenAPI Support
    • Parallel Inference
    • Adaptive Batching
    • Custom Inference Runtimes
    • Metrics
    • Deployment
      • Seldon Core
      • KServe
    • Streaming
  • Inference Runtimes
    • SKLearn
    • XGBoost
    • MLFlow
    • Spark MlLib
    • LightGBM
    • Catboost
    • Alibi-Detect
    • Alibi-Explain
    • HuggingFace
    • Custom
  • Reference
    • MLServer Settings
    • Model Settings
    • MLServer CLI
    • Python API
      • MLModel
      • Types
      • Codecs
      • Metrics
  • Examples
    • Serving Scikit-Learn models
    • Serving XGBoost models
    • Serving LightGBM models
    • Serving MLflow models
    • Serving a custom model
    • Serving Alibi-Detect models
    • Serving HuggingFace Transformer Models
    • Multi-Model Serving
    • Model Repository API
    • Content Type Decoding
    • Custom Conda environments in MLServer
    • Serving a custom model with JSON serialization
    • Serving models through Kafka
    • Streaming
    • Deploying a Custom Tensorflow Model with MLServer and Seldon Core
  • Changelog
Powered by GitBook
On this page
  • Usage
  • Content Types

Was this helpful?

Edit on GitHub
Export as PDF
  1. Inference Runtimes

MLFlow

This package provides a MLServer runtime compatible with MLflow models.

Usage

You can install the runtime, alongside mlserver, as:

pip install mlserver mlserver-mlflow

Content Types

The MLflow inference runtime introduces a new dict content type, which decodes an incoming V2 request as a dictionary of tensors. This is useful for certain MLflow-serialised models, which will expect that the model inputs are serialised in this format.

The `dict` content type can be _stacked_ with other content types, like
[`np`](../../docs/user-guide/content-type).
This allows the user to use a different set of content types to decode each of
the dict entries.
PreviousXGBoostNextSpark MlLib

Last updated 5 months ago

Was this helpful?