vllm.models.muse_spark.amd
¶
AMD ROCm implementation of Muse Spark.
Modules:
-
model–Muse Spark text model for AMD ROCm GPUs.
-
mono–FlyDSL decode monokernel for Muse Spark on MI355X (
--monokernel). -
multimodal–Muse Spark Omni for AMD ROCm GPUs: the shared Omni towers over the ROCm stack.
-
muse_moe_runner–ROCm MoE implementation for Muse Spark checkpoints.
-
ops–