Adrian Gonzalez-Martin
Adrian Gonzalez-Martin
It would be good to have a way to active / deactivate the idle inhibitor module programmatically (e.g. from a CLI). The use I have in mind is that whenever...
Part of #453 Explore what's the best way to remove the internal state from the `AnchorText` explainer.
Part of #453 Explore what's the best way to remove the internal state from the `AnchorTabular` explainer.
When leveraging Prometheus' pod annotations to configure the inference graph's scraping, you are not able to specify multiple ports to scrape. This poses some issues when considering multi-node inference graphs,...
Currently, the executor and the language wrapper use the OpenTracing libraries for tracing. However, that project is now moving into OpenTelemetry. OpenTelemetry claims to offer a standard protocol that works...
The OpenTelemetry project now has a `Collector` component, which can be seen as `fluentd` for traces. The collector offers a single (vendor agnostic) centralised point to gather traces from different...
Currently we only expose the `predict` method. However, some orchestrations frameworks like SC support the use of other "inference steps", like routing or aggregation. It would be good to explore...
MLServer now has built-in support to unpack and activate a [conda-pack](https://conda.github.io/conda-pack/) tarball. This feature could be leveraged to run the custom environment defined in the`conda.yaml` file usually present in MLflow...
To better support use cases where `mlserver` is used as a library, we shouldn't restrict the dependencies versions too much. Instead, we should look into adding some sort of lockfile...
MLServer uses FastAPI under the hood, which has some built-in functionality to generate and expose OpenAPI documentation on how to interact with the API. Additionally, the OpenAPI for the standard...