[Mini-vLLM] SM86-only cut: models, backends, Rust, tests

- Keep only Qwen3.5, Qwen3.5-MTP and Qwen4Exp model paths; trim the model registry to 41 entries
- Delete idle attention backends (MLA, FlashInfer, ROCm-only, quant backends) and dead GDN CuTeDSL kernels
- Remove inc/quark quant loaders; keep compressed-tensors and Marlin
- Drop the Rust frontend, pooling entrypoints, and idle test suites
- Rebuild kernels for sm_80;sm_86 and strip the extensions
- Verified: full import sweep clean (optional deps only) and a dummy-load Qwen3.5 engine run on RTX 3090 via the Triton GDN path

Co-authored-by: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
This commit is contained in:
2026-10-18 19:19:56 +07:00
co-authored by Hinata
parent c5ab42bc0f
commit e61dcb9143
2728 changed files with 159 additions and 950401 deletions
+4 -7
View File
@@ -11,7 +11,6 @@ from vllm.entrypoints.openai.models.serving import (
OpenAIModelRegistry,
OpenAIServingModels,
)
from vllm.entrypoints.pooling.typing import AnyPoolingRequest
from vllm.entrypoints.serve import create_error_response
from vllm.entrypoints.serve.engine.protocol import ErrorResponse
from vllm.entrypoints.serve.engine.typing import AnyRequest
@@ -48,7 +47,7 @@ class BaseServing:
async def _check_model(
self,
request: AnyRequest | AnyPoolingRequest,
request: AnyRequest,
) -> ErrorResponse | None:
if self.request_logger is not None:
self.request_logger.log_request_body(request)
@@ -141,7 +140,7 @@ class BaseServing:
return random_uuid() if default is None else default
def _get_message_types(self, request: AnyRequest | AnyPoolingRequest) -> set[str]:
def _get_message_types(self, request: AnyRequest) -> set[str]:
"""Retrieve the set of types from message content dicts up
until `_`; we use this to match potential multimodal data
with default per modality loras.
@@ -166,9 +165,7 @@ class BaseServing:
message_types.add(content_dict["type"].split("_")[0])
return message_types
def _get_active_default_mm_loras(
self, request: AnyRequest | AnyPoolingRequest
) -> LoRARequest | None:
def _get_active_default_mm_loras(self, request: AnyRequest) -> LoRARequest | None:
"""Determine if there are any active default multimodal loras."""
# TODO: Currently this is only enabled for chat completions
# to be better aligned with only being enabled for .generate
@@ -193,7 +190,7 @@ class BaseServing:
def _maybe_get_adapters(
self,
request: AnyRequest | AnyPoolingRequest,
request: AnyRequest,
supports_default_mm_loras: bool = False,
) -> LoRARequest | None:
if request.model in self.models.lora_requests: