[Edge][WP3] Add CPU OmniPlatform (AR + generation stages on vLLM's CPU backend)

* vllm_omni/platforms/cpu/platform.py: CPUOmniPlatform(OmniPlatform, CpuPlatform)
  selected when VLLM_TARGET_DEVICE=cpu or a +cpu vLLM wheel is installed
  (builtin plugin "cpu"; explicit CPU target short-circuits accelerator probes
  like vLLM does). NUMA-based memory queries, one logical device, no graph
  wrapper (sentinel type for isinstance checks), inductor opt-in via
  VLLM_OMNI_CPU_INDUCTOR=1, diffusion stages unsupported.
* vllm_omni/platforms/cpu/worker: CPUARWorker / CPUGenerationWorker compose
  vLLM's CPUWorker with the Omni mixin (XPU pattern); CPUARModelRunner /
  CPUGenerationModelRunner compose the Omni runners with CPUModelRunner
  (no graphs, no async omni output; sleep/wake return error ACKs).
* patch.py skips the CuMem patch on CPU-only torch; Qwen3-TTS speaker encoder
  and prompt-embeds dtype follow the model dtype (float32/float16 on CPU).
* vllm_omni/deploy/qwen3_tts.yaml: `platforms: cpu:` overlay (eager, no async
  scheduling, absolute kv_cache_memory_bytes, disjoint OMP thread ranges).
* requirements/cpu.txt (+psutil, intel-openmp), docs/getting_started/installation/cpu.md.

Measured (this x86 Xeon 8480C host, vllm 0.28.0+cpu, bf16, 12+4 threads):
Qwen3-TTS 0.6B CustomVoice end-to-end on CPU, 12 prompts: TTFA p50 292 ms,
RTF ~1.7 (25-frame chunk ~4.6 s), init 185 s. Result JSON + init timeline
committed under benchmarks/tts/edge_results/.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
tzhouamandClaude Fable 5.1 committed 2026-09-09 11:02:14 +08:00
1 parent 55192de984
commit 35e43b9983
18 files changed
+1740 -7

No files matched your search

+10 -2
View File
@@ -328,8 +328,8 @@ def _patch_chat_template_registry():
)
if "qwen3_omni_moe" not in _MODEL_TYPE_TO_CHAT_TEMPLATE_FALLBACK:
_MODEL_TYPE_TO_CHAT_TEMPLATE_FALLBACK["qwen3_omni_moe"] = (
lambda _: CHAT_TEMPLATES_DIR / "template_chatml.jinja"
_MODEL_TYPE_TO_CHAT_TEMPLATE_FALLBACK["qwen3_omni_moe"] = lambda _: (
CHAT_TEMPLATES_DIR / "template_chatml.jinja"
)
except ImportError:
pass
@@ -529,6 +529,14 @@ _patch_inductor_factorable_divisibility()
# (not a slot or C extension). If CuMemAllocator is rewritten in C/Cython,
# this monkey-patch will silently become a no-op.
def _patch_cumem_free_callback_cuda() -> None:
try:
import torch
if torch.version.cuda is None and getattr(torch.version, "hip", None) is None:
_PATCH_LOGGER.debug("[cumem-cuda] CPU-only torch; skipping patch")
return
except Exception:
pass
try:
from vllm.device_allocator.cumem import CuMemAllocator
except ImportError: