Run pretrained HuggingFace models — text embeddings, speech-to-text, and decoder LLMs — natively on the JVM on the raster typed-dispatch compiler. No Python, no ONNX runtime: weights load from safetensors, quantize to int8/int4, and run on raster's CPU int8-MAC kernels and Intel-GPU (Level Zero/OpenCL) programs.

• replikativ/pretrained-rstr
• cljdoc
• 0 Downloads
• 0 This Version

Leiningen/Boot

[org.replikativ/pretrained-rstr "0.1.1"]

Clojure CLI/deps.edn

org.replikativ/pretrained-rstr {:mvn/version "0.1.1"}

Gradle

implementation("org.replikativ:pretrained-rstr:0.1.1")

Maven

<dependency>
  <groupId>org.replikativ</groupId>
  <artifactId>pretrained-rstr</artifactId>
  <version>0.1.1</version>
</dependency>