SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2404.19124 · 2024

Accelerating Production LLMs with Combined Token/Embedding Speculators

arXiv · PDF · Open in the Atlas

Code that ran

5 of 9 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 5 failures of our instrument; 4 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/2404.19124.json · badge: /api/ran/2404.19124.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
foundation-model-stack/fms-fsdp canonical (classifier label)
classifier basis: census
5 ran · 4 no recorded run
FunctionStatusWhere it lives
causal_lm Ran
ran; our contract draft was wrong, not the code
foundation-model-stack/fms-fsdp/fms_fsdp/utils/dataloader_utils.py
code served (permissive licence) · get_code("3695fa5aa708290b")
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 2, dtype family int; 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 torch

torch.manual_seed(0)

# The function takes data_seq and prompt_len
# data_seq should be a sequence of integers that can be converted to a tensor
# prompt_len is an integer specifying how many tokens to ignore in loss

data_seq = [10, 20, 30, 40, 50]  # Simple sequence of integers

ARGS = {
    "data_seq": data_seq,
    "prompt_len": 1
}
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 tuple, ndim 2, dtype family int — 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
get_dummy_loader Ran
ran; our contract draft was wrong, not the code
foundation-model-stack/fms-fsdp/fms_fsdp/utils/dataloader_utils.py
code served (permissive licence) · get_code("6a5563e9d0bbdefa")
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 DataLoader. 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

torch.manual_seed(0)

class Config:
    def __init__(self):
        self.seq_length = 4
        self.vocab_size = 8
        self.batch_size = 2

cfg = Config()
rank = 0
world_size = 1

ARGS = {
    "cfg": cfg,
    "rank": rank,
    "world_size": world_size
}
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 DataLoader — a DataLoader output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataLoader 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
get_latest Ran
ran; our contract draft was wrong, not the code
foundation-model-stack/fms-fsdp/fms_fsdp/utils/checkpointing_utils.py
code served (permissive licence) · get_code("6aea1aac1421170f")
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 os
import tempfile
import shutil

# Create a temporary directory with some test files
temp_dir = tempfile.mkdtemp()
test_file_1 = os.path.join(temp_dir, "file1.txt")
test_file_2 = os.path.join(temp_dir, "file2.txt")
test_file_3 = os.path.join(temp_dir, "excluded_file.txt")

with open(test_file_1, 'w') as f:
    f.write("test1")
with open(test_file_2, 'w') as f:
    f.write("test2")
with open(test_file_3, 'w') as f:
    f.write("excluded")

# Define a qualifier function that excludes files with "excluded" in the name
def qualifier_fn(path):
    return "excluded" not in os.path.basename(path)

# Define arguments for the function
ARGS = {
    "targdir": temp_dir,
    "qualifier": qualifier_fn,
    "key": os.path.getctime
}
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
get_oldest Ran
ran; our contract draft was wrong, not the code
foundation-model-stack/fms-fsdp/fms_fsdp/utils/checkpointing_utils.py
code served (permissive licence) · get_code("d2cee2b98f8d1ad0")
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 os
import tempfile
import shutil

# Create a temporary directory with some test files
test_dir = tempfile.mkdtemp()

# Create some test files with different creation times
file1 = os.path.join(test_dir, "file1.txt")
file2 = os.path.join(test_dir, "file2.txt")
file3 = os.path.join(test_dir, "file3.txt")

with open(file1, 'w') as f:
    f.write("content1")
with open(file2, 'w') as f:
    f.write("content2")
with open(file3, 'w') as f:
    f.write("content3")

# Simple qualifier function that accepts all files
def simple_qualifier(path):
    return True

# Simple key function using os.path.getctime (default behavior)
def simple_key(path):
    return os.path.getctime(path)

ARGS = {
    "targdir": test_dir,
    "qualifier": simple_qualifier,
    "key": simple_key,
}
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
get_wrapper Ran
ran; our contract draft was wrong, not the code
foundation-model-stack/fms-fsdp/fms_fsdp/policies/wrapping.py
code served (permissive licence) · get_code("ed32994ac009007d")
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 partial. 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
import functools
from torch.distributed.fsdp.wrap import transformer_auto_wrap_policy

torch.manual_seed(0)

# Create a simple transformer block class to use as the block parameter
class SimpleTransformerBlock(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def forward(self, x):
        return self.linear(x)

# Instantiate a block
block = SimpleTransformerBlock()

ARGS = {
    "block": block
}
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 partial — a partial output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a partial 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
apply_fsdp_checkpointing Not yet run
unverified (harvested, no recorded run)
foundation-model-stack/fms-fsdp/fms_fsdp/policies/ac_handler.py
code served (permissive licence) · get_code("f5e2c63b0b777c26")
How it ran

no invocation recorded.

get_mixed_precision_policy Not yet run
unverified (harvested, no recorded run)
foundation-model-stack/fms-fsdp/fms_fsdp/utils/train_utils.py
code served (permissive licence) · get_code("ebc38f7d327d972f")
How it ran

no invocation recorded.

get_policies Not yet run
unverified (harvested, no recorded run)
foundation-model-stack/fms-fsdp/fms_fsdp/utils/train_utils.py
code served (permissive licence) · get_code("7cd33985533b9a09")
How it ran

no invocation recorded.

train Not yet run
unverified (harvested, no recorded run)
foundation-model-stack/fms-fsdp/fms_fsdp/utils/train_utils.py
code served (permissive licence) · get_code("aa7f3b5c0155bc7c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 9 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.

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

This technical report describes the design and training of novel speculative decoding draft models, for accelerating the inference speeds of large language models in a production environment. By conditioning draft predictions on both context vectors and sampled tokens, we can train our speculators to efficiently predict high-quality n-grams, which the base model then accepts or rejects. This allows us to effectively predict multiple tokens per inference forward pass, accelerating wall-clock inference speeds of highly optimized base model implementations by a factor of 2-3x. We explore these initial results and describe next steps for further improvements.

For agents

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

get_harvested_code_for_paper("2404.19124")
get_code_for_paper("2404.19124")
have("2404.19124")

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

curl https://syntology.ai/api/ran/2404.19124.json

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

[![Syntology run record](https://syntology.ai/api/ran/2404.19124.svg)](https://syntology.ai/paper/2404.19124)

Connect an agent — have() is free.