SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2512.09472 · ICML · 2025

WarmServe: Enabling One-for-Many GPU Prewarming for Multi-LLM Serving

Yong Zhang, Chen Sun, Xin Jin, Sheng Qi, Chiheng Lou, Rui Kang, Pengcheng Wang, Xuanzhe Liu

arXiv · PDF · Open in the Atlas

Code that ran

13 of 14 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 1 has no contract verdict; 11 failures of our instrument; 1 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2512.09472.json · badge: /api/ran/2512.09472.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
LLMServe/WarmServe canonical (classifier label)
classifier basis: census
13 ran · 1 no recorded run
FunctionStatusWhere it lives
apply_burstiness Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/trace-generator/trace_generator.py
code served (permissive licence) · get_code("47c3e4a0942c91ab")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a simple requests list with arrival times and additional data
# Format: (arrival_time, additional_data)
requests = [
    (0.0, "req_1"),
    (1.5, "req_2"),
    (3.2, "req_3"),
    (5.1, "req_4"),
    (7.8, "req_5"),
]

# Create a random number generator for reproducibility
rng = np.random.default_rng(42)

# Coefficient of variation values to test
cv = 0.8

ARGS = {
    "requests": requests,
    "cv": cv,
    "rng": rng,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 2, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
async_request_deepspeed_mii Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/vllm-0.6.3.post1/benchmarks/backend_request_func.py
code served (permissive licence) · get_code("f70b11daf69cb473")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned coroutine. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import asyncio
from dataclasses import dataclass, field
from typing import List, Optional
from tqdm.asyncio import tqdm

@dataclass
class RequestFuncInput:
    prompt: str
    api_url: str
    prompt_len: int
    output_len: int
    model: str
    best_of: int = 1
    logprobs: Optional[int] = None
    multi_modal_content: Optional[dict] = None
    ignore_eos: bool = False

@dataclass
class RequestFuncOutput:
    generated_text: str = ''
    success: bool = False
    latency: float = 0.0
    ttft: float = 0.0
    itl: List[float] = field(default_factory=list)
    prompt_len: int = 0
    error: str = ''

request_func_input = RequestFuncInput(
    prompt="What is machine learning?",
    api_url="http://localhost:8000/v1/completions",
    prompt_len=5,
    output_len=10,
    model="gpt-3.5-turbo",
    best_of=1,
    logprobs=None,
    multi_modal_content=None,
    ignore_eos=False
)

async def dummy_pbar():
    async for _ in tqdm(range(1)):
        pass
    return tqdm()

pbar = None

ARGS = {
    "request_func_input": request_func_input,
    "pbar": pbar
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
aiohttp, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type coroutine — a coroutine output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coroutine output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
async_request_tgi Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/vllm-0.6.3.post1/benchmarks/backend_request_func.py
code served (permissive licence) · get_code("5fe8fe4d65cc1519")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned coroutine. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import json
import sys
import time
import traceback
from dataclasses import dataclass, field
from typing import List, Optional, Union
from unittest.mock import AsyncMock, MagicMock, patch
from tqdm.asyncio import tqdm

@dataclass
class RequestFuncInput:
    prompt: str
    api_url: str
    prompt_len: int
    output_len: int
    model: str
    best_of: int = 1
    logprobs: Optional[int] = None
    multi_modal_content: Optional[dict] = None
    ignore_eos: bool = False

@dataclass
class RequestFuncOutput:
    generated_text: str = ''
    success: bool = False
    latency: float = 0.0
    ttft: float = 0.0
    itl: List[float] = field(default_factory=list)
    prompt_len: int = 0
    error: str = ''

# Create a minimal RequestFuncInput instance
request_func_input = RequestFuncInput(
    prompt="Hello, how are you?",
    api_url="http://localhost:8000/generate_stream",
    prompt_len=5,
    output_len=10,
    model="test-model",
    best_of=1,
    logprobs=None,
    multi_modal_content=None,
    ignore_eos=False
)

# Create a mock progress bar
mock_pbar = None

ARGS = {
    'request_func_input': request_func_input,
    'pbar': mock_pbar
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
aiohttp, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type coroutine — a coroutine output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coroutine output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
async_request_trt_llm Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/vllm-0.6.3.post1/benchmarks/backend_request_func.py
code served (permissive licence) · get_code("31b4648baccb4d32")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned coroutine. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import json
import sys
import time
import traceback
from dataclasses import dataclass, field
from typing import List, Optional, Union
from unittest.mock import AsyncMock, MagicMock, patch
import aiohttp
from tqdm.asyncio import tqdm

@dataclass
class RequestFuncInput:
    prompt: str
    api_url: str
    prompt_len: int
    output_len: int
    model: str
    best_of: int = 1
    logprobs: Optional[int] = None
    multi_modal_content: Optional[dict] = None
    ignore_eos: bool = False

@dataclass
class RequestFuncOutput:
    generated_text: str = ''
    success: bool = False
    latency: float = 0.0
    ttft: float = 0.0
    itl: List[float] = field(default_factory=list)
    prompt_len: int = 0
    error: str = ''

request_func_input = RequestFuncInput(
    prompt="This is a test prompt for the TRT-LLM API.",
    api_url="http://localhost:8000/generate_stream",
    prompt_len=10,
    output_len=50,
    model="test-model",
    best_of=1,
    logprobs=None,
    multi_modal_content=None,
    ignore_eos=False
)

pbar = None

ARGS = {
    "request_func_input": request_func_input,
    "pbar": pbar,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
aiohttp, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type coroutine — a coroutine output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coroutine output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
check_wheel_size Ran
ran, honoured the contract we drafted
LLMServe/WarmServe/vllm-0.6.3.post1/.buildkite/check-wheel-size.py
code served (permissive licence) · get_code("95e3362a392b24d6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import zipfile
import tempfile

# Create a temporary directory for testing
temp_dir = tempfile.mkdtemp()

# Create a test wheel file (it's just a zip file)
test_wheel_path = os.path.join(temp_dir, "test_package.whl")

# Create a small zip file with some test files
with zipfile.ZipFile(test_wheel_path, 'w') as z:
    z.writestr("file1.txt", "a" * (1024 * 100))  # 100 KB
    z.writestr("file2.txt", "b" * (1024 * 50))   # 50 KB
    z.writestr("file3.txt", "c" * (1024 * 25))   # 25 KB

ARGS = {
    "directory": temp_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_requests Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/trace-generator/request_generator.py
code served (permissive licence) · get_code("e92541d5932f8422")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import pickle
import tempfile

# Create a temporary pickle file with sample request data
temp_dir = tempfile.mkdtemp()
workload_path = os.path.join(temp_dir, "workload.pkl")

# Sample requests: (arrival_time, muxserve_id, model_id, input_tokens, output_tokens)
sample_requests = [
    (5.0, "model_1", "meta-llama/Llama-2-7b/tp1", 100, 50),
    (10.0, "model_2", "meta-llama/Llama-2-13b/tp2", 150, 75),
    (15.0, "model_1", "meta-llama/Llama-2-7b/tp1", 120, 60),
    (25.0, "model_3", "mistral/Mistral-7b/tp1", 200, 100),
    (1200.0, "model_1", "meta-llama/Llama-2-7b/tp1", 110, 55),
]

with open(workload_path, 'wb') as f:
    pickle.dump(sample_requests, f)

ARGS = {
    "workload_path": workload_path,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_type_from_path Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/trace-generator/trace_generator.py
code served (permissive licence) · get_code("73e947a7a2364546")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# The function takes a csv_path string parameter
# It checks if 'Conv' or 'conv' is in the path string
# and returns either 'conv' or 'code'

ARGS = {
    'csv_path': 'model_conv_weights.csv'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
repeat_and_sort_requests Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/vllm-0.6.3.post1/benchmarks/benchmark_prefix_caching.py
code served (permissive licence) · get_code("c6c44ee91bb762e9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import random
from typing import List, Tuple

# Set random seed for reproducibility
random.seed(0)

# Create sample request data
# Each request is a tuple of (request_name: str, priority: int, id: int)
requests = [
    ("request_A", 5, 101),
    ("request_B", 2, 102),
    ("request_C", 8, 103),
    ("request_D", 3, 104),
]

repeat_count = 2
sort = False

ARGS = {
    "requests": requests,
    "repeat_count": repeat_count,
    "sort": sort,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
run Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/vllm-0.6.3.post1/collect_env.py
code served (permissive licence) · get_code("6b73fe0480a96ad9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import locale
import subprocess
import sys

# The function 'run' takes a single parameter 'command'
# which can be either a string (shell=True) or a list (shell=False)
# It executes the command and returns (return_code, stdout, stderr)

# Valid arguments for the run function:
ARGS = {
    'command': ['echo', 'hello world']  # List form - will execute with shell=False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
run_and_parse_first_match Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/vllm-0.6.3.post1/collect_env.py
code served (permissive licence) · get_code("ce7a62e37fe78c0d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re

def mock_run_lambda(command):
    """Mock run_lambda that simulates running a shell command."""
    # Simulate successful execution with output containing a match
    if command == "test_command":
        return (0, "The version is 1.2.3 and other info", None)
    else:
        return (1, "", None)

def test_run_lambda_failure(command):
    """Mock run_lambda that returns non-zero exit code."""
    return (1, "error output", None)

def test_run_lambda_no_match(command):
    """Mock run_lambda that returns output with no regex match."""
    return (0, "output without version pattern", None)

# Define ARGS dict with all parameters
ARGS = {
    "run_lambda": mock_run_lambda,
    "command": "test_command",
    "regex": r"version is ([0-9.]+)"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
run_and_read_all Ran
ran; our contract draft was wrong, not the code
LLMServe/WarmServe/vllm-0.6.3.post1/collect_env.py
code served (permissive licence) · get_code("5ae630bfc3821b6e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

def create_run_lambda():
    """Create a mock run_lambda function that returns (rc, stdout, stderr)"""
    def mock_run_lambda(command):
        # Simulate successful command execution
        if "error" in command:
            return 1, "", "error occurred"
        else:
            return 0, "command output", ""
    return mock_run_lambda

ARGS = {
    "run_lambda": create_run_lambda(),
    "command": "echo 'test command'"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
sample_requests Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
LLMServe/WarmServe/vllm-0.6.3.post1/benchmarks/benchmark_prefix_caching.py
code served (permissive licence) · get_code("a8de224d74e65256")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import json
import random
import tempfile
import os
from typing import List, Optional, Tuple

class MockTokenizer:
    """Mock tokenizer that simulates tokenization behavior."""
    def __call__(self, text: str):
        # Simple mock: split by spaces and add some tokens
        tokens = text.split()
        # Return object with input_ids attribute
        class TokenizerOutput:
            def __init__(self, ids):
                self.input_ids = ids
        
        # Make token count depend on text length
        token_count = max(4, len(tokens) + len(text) // 10)
        return TokenizerOutput(list(range(token_count)))

# Create a temporary dataset file
temp_dir = tempfile.mkdtemp()
dataset_path = os.path.join(temp_dir, "dataset.json")

# Create sample dataset with conversations
dataset = [
    {
        "conversations": [
            {"value": "What is machine learning? " * 10},  # Long prompt
            {"value": "Machine learning is a subset of AI. " * 5}  # Medium completion
        ]
    },
    {
        "conversations": [
            {"value": "Tell me about Python. " * 8},
            {"value": "Python is a programming language. " * 4}
        ]
    },
    {
        "conversations": [
            {"value": "How does neural networks work? " * 12},
            {"value": "Neural networks are inspired by biology. " * 6}
        ]
    },
    {
        "conversations": [
            {"value": "Explain transformers. " * 9},
            {"value": "Transformers use attention mechanisms. " * 5}
        ]
    }
]

with open(dataset_path, 'w') as f:
    json.dump(dataset, f)

tokenizer = MockTokenizer()

ARGS = {
    "dataset_path": dataset_path,
    "num_requests": 3,
    "tokenizer": tokenizer,
    "input_length_range": (20, 200),
    "fixed_output_len": None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 2, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
single_load_generator Ran
ran on a synthesized input
LLMServe/WarmServe/trace-generator/request_generator.py
code served (permissive licence) · get_code("76a2d189cdb4c6cd")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned coroutine. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import sys

# Set environment variables that the function reads
os.environ['INIT_SEND'] = '0'
os.environ['WARM'] = '0'

# The function signature is:
# async def single_load_generator(add_mem_suffix: bool):
# 
# It takes a single boolean parameter add_mem_suffix

ARGS = {
    'add_mem_suffix': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
aiohttp — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type coroutine — a coroutine output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coroutine output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
find_cuda_init Not yet run
unverified (harvested, no recorded run)
LLMServe/WarmServe/vllm-0.6.3.post1/find_cuda_init.py
code served (permissive licence) · get_code("6a9382e5d439d948")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 6 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 6 later) and 8 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
run_and_parse_first_match 14 other papers
10 dated earlier, 3 later, 1 same year, order not determinable
2109.04003 (2021-09)
Improving Deep Metric Learning by Divide and Conquer
2109.08267 (2021-09)
CompilerGym: Robust, Performant Compiler Optimization Environments for AI Research
2202.04053 (2022-02)
DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Models
and 11 more not named here
run_and_read_all 14 other papers
10 dated earlier, 3 later, 1 same year, order not determinable
2109.04003 (2021-09)
Improving Deep Metric Learning by Divide and Conquer
2109.08267 (2021-09)
CompilerGym: Robust, Performant Compiler Optimization Environments for AI Research
2202.04053 (2022-02)
DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Models
and 11 more not named here
run 6 other papers
4 dated earlier, 1 later, 1 same year, order not determinable
2405.16444 (2024-05)
CacheBlend: Fast Large Language Model Serving for RAG with Cached Knowledge Fusion
2406.19215 (2024-06)
SeaKR: Self-aware Knowledge Retrieval for Adaptive Retrieval Augmented Generation
2410.00428 (2024-10)
LayerKV: Optimizing Large Language Model Serving with Layer-wise KV Cache Management
and 3 more not named here
check_wheel_size 4 other papers
3 later, 1 same year, order not determinable
2025.emnlp-main.240 (2025 (year only))
QSPEC: Speculative Decoding with Complementary Quantization Schemes
2601.20309 (2026-01)
SuperInfer: SLO-Aware Rotary Scheduling and Memory Management for LLM Inference on Superchips
2602.07223 (2026-02)
Vegas: Self-Speculative Decoding with Verification-Guided Sparse Attention
and 1 more not named here
async_request_deepspeed_mii 2 other papers
1 later, 1 same year, order not determinable
2025.emnlp-main.240 (2025 (year only))
QSPEC: Speculative Decoding with Complementary Quantization Schemes
2601.20309 (2026-01)
SuperInfer: SLO-Aware Rotary Scheduling and Memory Management for LLM Inference on Superchips
find_cuda_init 2 other papers
1 later, 1 same year, order not determinable
2025.emnlp-main.240 (2025 (year only))
QSPEC: Speculative Decoding with Complementary Quantization Schemes
2601.20309 (2026-01)
SuperInfer: SLO-Aware Rotary Scheduling and Memory Management for LLM Inference on Superchips

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Deploying multiple models within shared GPU clusters is a key strategy to improve resource efficiency in large language model (LLM) serving. Existing multi-LLM serving systems improve GPU utilization at the cost of degraded inference performance, particularly time-to-first-token (TTFT). We attribute this degradation to the lack of awareness regarding future workload characteristics. In contrast, recent analyses have shown the strong periodicity and long-term predictability of real-world LLM serving workloads. In this paper, we propose one-for-many GPU prewarming, which proactively loads parameters from multiple models onto GPUs based on workload forecasts. These prewarmed weights enable the system to promptly instantiate serving instances upon encountering request bursts. We design and implement WarmServe, a multi-LLM serving system incorporating three key techniques: (1) a model placement algorithm that optimizes prewarming decisions to minimize cross-model prewarming interference, (2) a KV cache reservation strategy that repurposes idle KV cache space on running GPUs for prewarming new models, and (3) an efficient GPU memory switching mechanism for tensor management. Evaluation on real-world datasets shows that WarmServe reduces tail TTFT by up to 50.8× compared to the state-of-the-art autoscaling-based system, while supporting up to 2.5× higher request throughput than the GPUsharing system.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2512.09472")
get_code_for_paper("2512.09472")
have("2512.09472")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2512.09472.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2512.09472.svg)](https://syntology.ai/paper/2512.09472)

Connect an agent — have() is free.