[Mini-VLLM] Cut non-OpenAI surfaces, dead kernels, Mistral stack

- Remove Anthropic, Cohere, SageMaker, decisions, structured-decisions,
  rl, render-launcher and batch entrypoints; keep scale-out token routes
- Drop the Swagger static bundle and the /docs, /redoc routes
- Delete device-keyed tuning configs for fused_moe, quant utils, mamba
  select-state and helion; no sm86 entry ever existed for any of them
- Remove ten unused rope variants plus the fbgemm_fp8 and fp_quant
  quantizers and their config, metrics and loader plumbing
- Cut the CPU device build: csrc/cpu, cpu_extension.cmake, cpu
  requirements and the setup.py cpu target path
- Delete 29 environment variables nothing reads and 14 chat-template
  fallbacks for models this fork no longer registers
- Remove nine dead CUDA ops (deepseek_v4, kimi_k3, minimax, muse_spark,
  dsv3 gemm) with their kernels, bindings, wrappers and CMake entries
- Delete the connector test suites that targeted removed connectors
- Cut the Mistral tokenizer and parser stack, its tests, and the
  mistral_common and ijson dependency pins
- Restore the Qwen4Exp registry rows and nvidia/mtp.py that an earlier
  cut deleted while keeping the model tree

Co-authored-by: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>
This commit is contained in:
2026-10-19 17:10:56 +07:00
co-authored by Hinata
parent bd70a61c20
commit 2066d3143f
1086 changed files with 122 additions and 291919 deletions
-64
View File
@@ -23,7 +23,6 @@ from multiprocessing import Process, get_context
from pathlib import Path
from typing import TYPE_CHECKING, Any, Literal, cast
import anthropic
import cloudpickle
import httpx
import openai
@@ -717,23 +716,6 @@ class RemoteVLLMServer:
**kwargs,
)
def get_client_anthropic(self, **kwargs):
if "timeout" not in kwargs:
kwargs["timeout"] = 600
return anthropic.Anthropic(
base_url=self.url_for(),
api_key=self.DUMMY_API_KEY,
max_retries=0,
**kwargs,
)
def get_async_client_anthropic(self, **kwargs):
if "timeout" not in kwargs:
kwargs["timeout"] = 600
return anthropic.AsyncAnthropic(
base_url=self.url_for(), api_key=self.DUMMY_API_KEY, max_retries=0, **kwargs
)
class RemoteOpenAIServer(RemoteVLLMServer):
"""Launches ``vllm serve`` for testing OpenAI-compatible endpoints."""
@@ -773,52 +755,6 @@ class RemoteOpenAIServer(RemoteVLLMServer):
)
class RemoteLaunchRenderServer(RemoteVLLMServer):
"""Launches ``vllm launch render`` for GPU-less serving tests."""
def _create_cli_subcommand(self):
return ServeSubcommand()
def _start_server(
self, model: str, vllm_serve_args: list[str], env_dict: dict[str, str] | None
) -> None:
env = os.environ.copy()
env["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
if env_dict is not None:
env.update(env_dict)
_sanitize_pythonpath_env(env)
serve_cmd = ["vllm", "launch", "render", model, *vllm_serve_args]
redacted_serve_cmd = _redact_sensitive_cli_args(serve_cmd)
print(
f"Launching RemoteLaunchRenderServer with: {' '.join(redacted_serve_cmd)}"
)
self.proc: subprocess.Popen = subprocess.Popen(
serve_cmd,
env=env,
stdout=sys.stdout,
stderr=sys.stderr,
start_new_session=True,
)
def _pre_download_model(self, model: str, args) -> None:
"""Download only the tokenizer files (no model weights needed)."""
is_local = os.path.isdir(model)
if not is_local:
engine_args = AsyncEngineArgs.from_cli_args(args)
model_config = engine_args.create_model_config()
get_tokenizer(
model_config.tokenizer,
tokenizer_mode=model_config.tokenizer_mode,
trust_remote_code=model_config.trust_remote_code,
revision=model_config.tokenizer_revision,
)
def _wait_for_gpu_memory_release(
self, timeout: float = 30.0, log_interval: float = 10.0
):
pass # No GPU used
class RemoteOpenAIServerCustom(RemoteOpenAIServer):
"""Launch test server with custom child process."""