2.8 KiB
2.8 KiB
User guides
How-to guides for deploying, scaling, and operating Ray Serve LLM. If you are new, start with the {doc}Quickstart <../quick-start>, then come back here to go deeper.
Configure and deploy
- {doc}
Configuration reference <configuration>: everyLLMConfigfield, from model loading and engine kwargs to accelerators, placement, and deployment options. - {doc}
Deployment initialization <deployment-initialization>: speed up model loading and replica startup with caching, streaming load formats, and initialization callbacks. - {doc}
Multi-LoRA deployment <multi-lora>: serve many LoRA adapters on a shared base model with runtime switching and an LRU cache.
Scale across GPUs and nodes
- {doc}
Cross-node parallelism <cross-node-parallelism>: distribute a model across GPUs and nodes with tensor and pipeline parallelism and placement groups. - {doc}
Data parallel attention <data-parallel-attention>: replicate the model into coordinated data-parallel groups to raise throughput, especially for MoE models. - {doc}
Fractional GPU serving <fractional-gpu>: pack multiple small-model replicas onto a single GPU.
Optimize latency and throughput
- {doc}
Prefill/decode disaggregation <prefill-decode>: split prompt processing and token generation onto separate replicas to tune each independently. - {doc}
KV cache offloading <kv-cache-offloading>: extend KV cache capacity with LMCache and tiered storage backends. - {doc}
Prefix-aware routing <prefix-aware-routing>: route requests to replicas that already hold a matching prefix to maximize cache hits. - {doc}
Direct streaming <direct-streaming>: bypass the ingress when streaming tokens to cut per-token latency.
Choose an engine
- {doc}
vLLM compatibility <vllm-compatibility>: use vLLM features such as embeddings, structured outputs, vision, and reasoning through Ray Serve LLM. - {doc}
Custom vLLM models <custom-vllm>: serve an out-of-tree architecture with a vLLM plugin, using a Qwen3 reward model as the example. - {doc}
SGLang integration <sglang>: run SGLang as the inference engine instead of vLLM.
Operate in production
- {doc}
Observability and monitoring <observability>: engine and request metrics, Grafana dashboards, and Prometheus integration.
:hidden:
:maxdepth: 1
Configuration reference <configuration>
Deployment initialization <deployment-initialization>
Multi-LoRA deployment <multi-lora>
Cross-node parallelism <cross-node-parallelism>
Data parallel attention <data-parallel-attention>
Fractional GPU serving <fractional-gpu>
Prefill/decode disaggregation <prefill-decode>
KV cache offloading <kv-cache-offloading>
Prefix-aware routing <prefix-aware-routing>
Direct streaming <direct-streaming>
vLLM compatibility <vllm-compatibility>
Custom vLLM models <custom-vllm>
SGLang integration <sglang>
Observability and monitoring <observability>