https://hexfusion.io/tags/llm-d/Recent content in Llm-D on hexfusion - Sam BatscheletHugoen-usSat, 21 Mar 2026 00:00:00 +0000https://hexfusion.io/posts/gie-1.4-framework-release/Sat, 21 Mar 2026 00:00:00 +0000https://hexfusion.io/posts/gie-1.4-framework-release/Gateway API Inference Extension v1.4 landed with 101 commits from 54 contributors. The headline isn’t a single feature, it’s that GIE became a real plugin framework. Here’s what changed and why it matters if you’re building on top of it.https://hexfusion.io/posts/lora-go-training-pt2/Thu, 19 Mar 2026 00:00:00 +0000https://hexfusion.io/posts/lora-go-training-pt2/The v2 adapter trained overnight on 41k samples. Loss 0.918, accuracy 82.7%. I loaded it into vLLM and ran the same prompts. Here’s what came out.https://hexfusion.io/posts/lora-go-training-pt1/Wed, 18 Mar 2026 00:00:00 +0000https://hexfusion.io/posts/lora-go-training-pt1/I trained a LoRA adapter on 41k Go code examples from the Kubernetes and etcd source trees. The first run produced 600 tab characters. Here’s what I learned.https://hexfusion.io/posts/dranet-bare-metal-rdma/Tue, 17 Mar 2026 00:00:00 +0000https://hexfusion.io/posts/dranet-bare-metal-rdma/hostNetwork is the default recommendation for RDMA in Kubernetes. It breaks disaggregated inference. DRANet replaces it with DRA-based NIC assignment and fixes the problem cleanly.https://hexfusion.io/posts/disaggregated-pd-consumer-gpus/Sat, 14 Mar 2026 00:00:00 +0000https://hexfusion.io/posts/disaggregated-pd-consumer-gpus/Running llm-d’s disaggregated prefill/decode architecture across an RTX 3060 and a Tesla T4 connected by 25GbE RDMA. What worked, what broke, and what I learned about KV cache transfer at the edge of what consumer hardware can do.