# SPDX-License-Identifier: Apache-2.0 # SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project """Spark-X2.5 as an Omni pipeline: one autoregressive text stage. Spark's model definition has been in ``vllm_omni.model_executor.models`` for a while, but only as a vLLM architecture -- there was no pipeline, so ``Omni(model=)`` had nothing to resolve and the model was reachable only through plain ``vllm.LLM``. That is the gap M0 closes: the text mode has to run *through* the Omni stage runtime, or none of the stage lifecycle, admission or cancellation work applies to it. There is exactly one stage, and that is the design, not a simplification to be undone later. The proposal's first decision is that an autoregressive session stays on one backend by default: prefill and decode share the KV cache, and a per-token boundary between devices buys nothing and costs a synchronization. Spark's tokenizer, chat template and sampling are model-adapter concerns that run in-process around this stage; they are not stages. """ from vllm_omni.config.stage_config import ( PipelineConfig, StageExecutionType, StagePipelineConfig, ) SPARK2_5_PIPELINE = PipelineConfig( model_type="spark2_5", model_arch="Spark2_5ForCausalLM", hf_architectures=("Spark2_5ForCausalLM",), stages=( StagePipelineConfig( stage_id=0, model_stage="spark2_5", execution_type=StageExecutionType.LLM_AR, input_sources=(), final_output=True, final_output_type="text", owns_tokenizer=True, engine_output_type="text", model_arch="Spark2_5ForCausalLM", ), ), ) # A separate whole-session adapter for a pinned GGUF artifact. The complete # request path reuses Omni's bounded graph-stage control plane, while llama.cpp # owns the model weights, KV and sampling within the single backend process. SPARK2_5_GGUF_TEXT_PIPELINE = PipelineConfig( model_type="spark2_5_gguf_text", stages=( StagePipelineConfig( stage_id=0, model_stage="text", execution_type=StageExecutionType.GRAPH, final_output=True, final_output_type="text", ), ), )