Search Results for "rule based expert system"

EPLB is DeepSeek’s open implementation of a load balancing algorithm designed for expert parallelism (EP) settings in MoE architectures. In EP, different “experts” are mapped to different GPUs or nodes, so load imbalance becomes a performance bottleneck if certain experts are invoked much more often. EPLB solves this by duplicating heavily used experts (redundancy) and then placing those duplicates across GPUs to even out computational load. It uses policies like hierarchical load balancing...

Grafana OSS is the leading open-source platform for visualization and observability. It enables teams to query, visualize, alert on, and explore telemetry data from multiple sources in a single interface. With support for 100+ data source plugins—including Prometheus, Loki, Elasticsearch, InfluxDB, SQL/NoSQL databases, and OpenTelemetry—Grafana helps teams correlate metrics, logs, and traces across applications and infrastructure. Users can build interactive dashboards with rich...

RadicalSpam is a free and open source package distributed under GPL v2, including products such as Postfix, SpamAssassin Amavisd-new, Clamav, Razor, DCC, Postgrey, Bind; providing a secure SMTP relay, ready to use with linux and docker environement. More information : http://www.radical-spam.org

Scripts and programs to make easy expert users life (basically POSIX users)

This has a Python ICAP (Internet Content Adaptation Protocol) server and IRML parser. Using this, a web proxy can do rule-based adaptation of content before delivery to clients. Has a Python proxylet API and squid-icap-client code too.