[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
@@ -2,7 +2,6 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import multiprocessing
import os
import socket
import threading
import time
@@ -410,198 +409,3 @@ def test_external_process_monitoring(api_server_args):
mock_coordinator.shutdown()
time.sleep(0.2)
WORKER_RUNTIME_SECONDS = prev_worker_runtime
def test_rust_frontend_launch_log_redacts_credentials(monkeypatch):
"""The Rust frontend command carries every non-default arg as JSON.
Credentials must not reach the log line."""
import subprocess as subprocess_mod
from vllm.entrypoints.launchers.cli_args import make_arg_parser
from vllm.utils.argparse_utils import FlexibleArgumentParser
from vllm.v1.utils import RustFrontendProcessManager
hf_token = "hf_TESTTOKEN123"
api_key = "sk-TESTKEY456"
args = make_arg_parser(FlexibleArgumentParser()).parse_args(
["--model", "org/model", "--hf-token", hf_token, "--api-key", api_key]
)
class _FakeProc:
pid = 4321
returncode = 0
def poll(self):
return 0
def wait(self, timeout=None):
return 0
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
input_listener = make_zmq_listener("tcp://127.0.0.1:0", zmq.ROUTER)
output_listener = make_zmq_listener("tcp://127.0.0.1:0", zmq.PULL)
try:
monkeypatch.setattr(subprocess_mod, "Popen", lambda *a, **kw: _FakeProc())
with patch("vllm.v1.utils.logger.info") as log_info:
RustFrontendProcessManager(
binary_path="/nonexistent/vllm-rs",
sock=sock,
args=args,
input_listener=input_listener,
output_listener=output_listener,
engine_start_index=0,
engine_count=1,
data_parallel_size=1,
)
finally:
sock.close()
log_format, *log_args = log_info.call_args.args
message = log_format % tuple(log_args)
assert "Launching Rust frontend:" in message
assert hf_token not in message
assert api_key not in message
assert '"hf_token": "***"' in message
assert '"api_key": "***"' in message
def test_rust_frontend_uses_config_model_as_model_tag(monkeypatch, caplog, tmp_path):
"""Config-only model selection must be forwarded to the Rust frontend."""
import subprocess as subprocess_mod
from vllm.entrypoints.launchers.cli_args import make_arg_parser
from vllm.utils.argparse_utils import FlexibleArgumentParser
from vllm.v1.utils import RustFrontendProcessManager
config_path = tmp_path / "config.yml"
config_path.write_text("model: org/model\n")
args = make_arg_parser(FlexibleArgumentParser()).parse_args(
["--config", str(config_path)]
)
assert args.model == "org/model"
assert args.model_tag is None
class _FakeProc:
pid = 4321
returncode = 0
def poll(self):
return 0
def wait(self, timeout=None):
return 0
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
input_listener = make_zmq_listener("tcp://127.0.0.1:0", zmq.ROUTER)
output_listener = make_zmq_listener("tcp://127.0.0.1:0", zmq.PULL)
try:
monkeypatch.setattr(subprocess_mod, "Popen", lambda *a, **kw: _FakeProc())
with caplog.at_level("INFO", logger="vllm.v1.utils"):
RustFrontendProcessManager(
binary_path="/nonexistent/vllm-rs",
sock=sock,
args=args,
input_listener=input_listener,
output_listener=output_listener,
engine_start_index=0,
engine_count=1,
data_parallel_size=1,
)
finally:
sock.close()
assert '"model_tag": "org/model"' in caplog.text
def test_rust_frontend_inherits_grpc_listener(monkeypatch):
"""The Python-bound gRPC socket is passed to Rust like the HTTP one."""
import subprocess as subprocess_mod
from vllm.entrypoints.launchers.cli_args import make_arg_parser
from vllm.utils.argparse_utils import FlexibleArgumentParser
from vllm.v1.utils import RustFrontendProcessManager
args = make_arg_parser(FlexibleArgumentParser()).parse_args(
["--model", "org/model", "--grpc-port", "50051"]
)
class _FakeProc:
pid = 4321
returncode = 0
def poll(self):
return 0
def wait(self, timeout=None):
return 0
popen_calls = []
def fake_popen(cmd, **kwargs):
popen_calls.append((cmd, kwargs))
return _FakeProc()
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
grpc_sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
input_listener = make_zmq_listener("tcp://127.0.0.1:0", zmq.ROUTER)
output_listener = make_zmq_listener("tcp://127.0.0.1:0", zmq.PULL)
input_fd = input_listener.socket.fileno()
output_fd = output_listener.socket.fileno()
try:
monkeypatch.setattr(subprocess_mod, "Popen", fake_popen)
RustFrontendProcessManager(
binary_path="/nonexistent/vllm-rs",
sock=sock,
args=args,
input_listener=input_listener,
output_listener=output_listener,
engine_start_index=0,
engine_count=1,
data_parallel_size=1,
grpc_sock=grpc_sock,
)
[(cmd, kwargs)] = popen_calls
grpc_fd = grpc_sock.fileno()
flag_index = cmd.index("--grpc-listen-fd")
assert cmd[flag_index + 1] == str(grpc_fd)
assert list(kwargs["pass_fds"]) == [
sock.fileno(),
input_fd,
output_fd,
grpc_fd,
]
finally:
sock.close()
grpc_sock.close()
def test_rust_frontend_spawn_failure_cleans_up_listeners(tmp_path):
"""No child owns the listeners when Popen fails, so the parent removes them."""
from vllm.entrypoints.launchers.cli_args import make_arg_parser
from vllm.utils.argparse_utils import FlexibleArgumentParser
from vllm.v1.utils import RustFrontendProcessManager
args = make_arg_parser(FlexibleArgumentParser()).parse_args(
["--model", "org/model"]
)
input_listener = make_zmq_listener(f"ipc://{tmp_path}/input.sock", zmq.ROUTER)
output_listener = make_zmq_listener(f"ipc://{tmp_path}/output.sock", zmq.PULL)
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
try:
with pytest.raises(FileNotFoundError):
RustFrontendProcessManager(
binary_path=str(tmp_path / "missing-vllm-rs"),
sock=sock,
args=args,
input_listener=input_listener,
output_listener=output_listener,
engine_start_index=0,
engine_count=1,
data_parallel_size=1,
)
finally:
sock.close()
for listener in (input_listener, output_listener):
assert listener.socket.fileno() == -1
assert not os.path.exists(listener.address.removeprefix("ipc://"))