2 repos
alexellis/glm-5.3-flash-4x-dgx-spark-switchless
GLM-5.3-Flash (NVFP4) at TP4 across 4x DGX Spark via a switchless RoCE ring + DFlash2 —…
87
10 commits
leyten/shard
Pipeline-parallel LLM inference across GPUs on separate machines.
456
363 commits