SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.02597 · 2026

GPUTOK: GPU Accelerated Byte Level BPE Tokenization

Data Science, Venu Gopal, Kadamba In, Kanishkha Jaisankar, M In

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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 failures of our instrument; 4 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/2603.02597.json · badge: /api/ran/2603.02597.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
venugopalkadamba/gpu-tokenizer canonical (classifier label)
classifier basis: census
5 ran · 4 no recorded run
FunctionStatusWhere it lives
load_gpu_tokens Ran
ran; our contract draft was wrong, not the code
venugopalkadamba/gpu-tokenizer/compare_tokenizers.py
code served (permissive licence) · get_code("2a602d58c40245f0")
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. 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.

from pathlib import Path
import tempfile

# Create a temporary directory and file with sample token data
temp_dir = tempfile.TemporaryDirectory()
temp_path = Path(temp_dir.name) / "gpu_tokens.txt"

# Write sample token data to the file
with temp_path.open('w', encoding='utf-8') as f:
    f.write("token1 token2 token3\n")
    f.write("token4 token5\n")
    f.write("\n")
    f.write("token6\n")

ARGS = {
    'path': temp_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 list — a list 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
load_lines Ran
ran; our contract draft was wrong, not the code
venugopalkadamba/gpu-tokenizer/compare_tokenizers.py
code served (permissive licence) · get_code("a4f33563585ce421")
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 tempfile
from pathlib import Path

# Create a temporary file with some test lines
with tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt', encoding='utf-8') as f:
    f.write("line one\n")
    f.write("line two\n")
    f.write("line three\n")
    temp_path = f.name

path_obj = Path(temp_path)

ARGS = {
    "path": path_obj
}
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
load_text Ran
ran; our contract draft was wrong, not the code
venugopalkadamba/gpu-tokenizer/compare_tokenizers.py
code served (permissive licence) · get_code("0448325414a91eb0")
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.

from pathlib import Path
import tempfile
import os

# Create a temporary file with some text content
temp_dir = tempfile.mkdtemp()
temp_file_path = Path(temp_dir) / "test.txt"
with open(temp_file_path, 'w', encoding='utf-8') as f:
    f.write("Hello, this is a test file.\nIt contains multiple lines.\n")

ARGS = {
    "path": temp_file_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 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_blockbpe_binary_on_prompts Ran
ran; our contract draft was wrong, not the code
venugopalkadamba/gpu-tokenizer/benchmark_tokenizer_scaling.py
code served (permissive licence) · get_code("3565560c2c6a0234")
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 tempfile
from pathlib import Path

# Create minimal prompts for testing
prompts = [
    "Hello world this is a test prompt for blockbpe tokenization",
    "Another sample prompt with some text to tokenize"
]

# Create a temporary directory and a dummy executable path
# (the function will handle the case where the binary doesn't exist)
temp_dir = tempfile.mkdtemp()
exe_path = Path(temp_dir) / "blockbpe_binary"

# Create the argument dictionary
ARGS = {
    "prompts": prompts,
    "exe_path": exe_path,
    "chunk_tokens": 512,
    "timeout_s": 300
}
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
tokenize_with_hf Ran
ran, honoured the contract we drafted
venugopalkadamba/gpu-tokenizer/compare_generation_quality.py
code served (permissive licence) · get_code("ad60d32392fdaeae")
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 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
import numpy as np
from typing import List

class MockTokenizer:
    """Mock HuggingFace tokenizer with encode method."""
    def encode(self, text: str, add_special_tokens: bool = False) -> List[int]:
        # Simple mock: return token IDs based on text length and character codes
        return [ord(c) % 1000 for c in text]

ARGS = {
    "text": "Hello world",
    "tokenizer": MockTokenizer(),
}
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 list, ndim 1, dtype family int — 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
build_gpu_extension Not yet run
unverified (harvested, no recorded run)
venugopalkadamba/gpu-tokenizer/benchmark_tokenizer_scaling.py
code served (permissive licence) · get_code("86612a0fed018432")
How it ran

no invocation recorded.

load_wikitext_stream Not yet run
unverified (harvested, no recorded run)
venugopalkadamba/gpu-tokenizer/benchmark_tokenizer_scaling.py
code served (permissive licence) · get_code("a10bce24d5c1e2fb")
How it ran

no invocation recorded.

map_token_strings_to_ids Not yet run
unverified (harvested, no recorded run)
venugopalkadamba/gpu-tokenizer/compare_generation_quality.py
code served (permissive licence) · get_code("9695cfdddf9e289e")
How it ran

no invocation recorded.

tokenize_with_tiktoken Not yet run
unverified (harvested, no recorded run)
venugopalkadamba/gpu-tokenizer/compare_generation_quality.py
code served (permissive licence) · get_code("175d6803059e0fd7")
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

As large language models move toward million-token context windows, CPU tokenizers become a major slowdown because they process text one step at a time while powerful GPUs sit unused. We built a GPU-based byte-level BPE tokenizer that follows GPT-2's merge rules. It includes a basic BlockBPE-style kernel and a faster, optimized version that uses cuCollections static map, CUB reductions, and a pybind11 interface for Python. On WikiText103 sequences up to 131k tokens, the optimized GPU tokenizer produces the same tokens as a CPU version and, for the longest inputs, is about 1.7× faster than tiktoken and about 7.6× faster than the HuggingFace GPT-2 tokenizer. Nsight profiling shows that 70-80% of CUDA API time goes to memory allocation, so adding memory pooling should give the biggest speed boost next. Tests on generation tasks using WikiText103 prompts show that our GPU tokenizer's outputs stay within about one percentage point of tiktoken and HuggingFace GPT-2 on similarity and overlap metrics, meaning it keeps output quality while making longcontext inference more practical.

For agents

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

get_harvested_code_for_paper("2603.02597")
get_code_for_paper("2603.02597")
have("2603.02597")

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

curl https://syntology.ai/api/ran/2603.02597.json

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

[![Syntology run record](https://syntology.ai/api/ran/2603.02597.svg)](https://syntology.ai/paper/2603.02597)

Connect an agent — have() is free.