SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2010.06192 · 2020

Revisiting BFloat16 Training

arXiv · PDF · Open in the Atlas

Code that ran

6 of 8 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 5 have no contract verdict; 1 failure of our instrument; 2 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/2010.06192.json · badge: /api/ran/2010.06192.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
eole-nlp/eole community (classifier label)
classifier basis not recorded on the link
6 ran · 2 no recorded run
FunctionStatusWhere it lives
extract_chunks Ran
ran on a synthesized input
eole-nlp/eole/apps/eole-translator.py
code served (permissive licence) · get_code("0a7d3b2dea77a6c8")
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 re

def extract_chunks(text):
    parts = re.split('(\\n+)', text)
    sentences, breaks = ([], [])
    for part in parts:
        if part.strip() == '':
            breaks.append(part)
        else:
            sentences.append(part.strip())
    return ([s for s in sentences if s], breaks)

# Setup code
ARGS = {
    'text': "Hello world\n\nThis is a test\n\n\nMultiple breaks"
}
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_base_url Ran
ran on a synthesized input
eole-nlp/eole/apps/eole-translator.py
code served (permissive licence) · get_code("5482b0753733ae86")
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 urllib.parse import urlparse, urlunparse

def get_base_url(api_url):
    parsed = urlparse(api_url)
    return urlunparse((parsed.scheme, parsed.netloc, '', '', '', ''))

ARGS = {
    "api_url": "https://api.example.com/v1/users/123?key=value#section"
}
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
hf_cache_dir Ran
ran on a synthesized input
eole-nlp/eole/eole/models/hf_resolver.py
code served (permissive licence) · get_code("20b14ae6f5d3a549")
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

def hf_cache_dir(model_id):
    slug = model_id.replace('/', '--').replace('\\', '--')
    return os.path.join(os.path.expanduser('~'), '.cache', 'eole', 'hf', slug)

ARGS = {
    'model_id': 'meta-llama/Llama-2-7b'
}
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
is_hf_model_id Ran
ran on a synthesized input
eole-nlp/eole/eole/models/hf_resolver.py
code served (permissive licence) · get_code("9ff19f37657a0682")
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 bool, ndim 0, dtype family float, 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

def is_hf_model_id(model_path):
    """Return True if model_path should be treated as a Hugging Face model ID.

    A path is treated as an HF model ID when it is relative, contains a slash,
    and does not exist on the local filesystem. This avoids misclassifying
    absolute or existing local model directories as Hub IDs.
    """
    s = str(model_path)
    if os.path.isabs(s) or s.startswith('.'):
        return False
    return '/' in s and (not os.path.exists(s))


# Test cases for is_hf_model_id
# The function checks:
# 1. If path is absolute -> False
# 2. If path starts with '.' -> False
# 3. If path contains '/' AND doesn't exist locally -> True
# 4. Otherwise -> False

# Valid HF model ID: relative path, contains slash, doesn't exist
ARGS = {
    'model_path': 'organization/model-name'
}
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 bool, ndim 0, dtype family float, 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
pad_at_dim Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
eole-nlp/eole/eole/modules/alibi_position_bias.py
code served (permissive licence) · get_code("e62e48a00ed06f63")
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 Tensor, ndim 2, dtype family float, shape (2, 7) 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 torch
import torch.nn.functional as F

torch.manual_seed(0)

# Create a small tensor for testing
# Shape: (2, 4) - batch size 2, feature dim 4
t = torch.randn(2, 4)

# Define padding as (left, right) tuple for the dimension
pad = (1, 2)

# Dimension to pad (-1 means last dimension)
dim = -1

# Padding value
value = 0.0

ARGS = {
    "t": t,
    "pad": pad,
    "dim": dim,
    "value": value,
}
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 Tensor, ndim 2, dtype family float, shape (2, 7) — shape (2, 7) 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
rebuild_text Ran
ran on a synthesized input
eole-nlp/eole/apps/eole-translator.py
code served (permissive licence) · get_code("4ed31e70485394dc")
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

def rebuild_text(translated_sentences, breaks):
    result = []
    s_idx = b_idx = 0
    while s_idx < len(translated_sentences) or b_idx < len(breaks):
        if s_idx < len(translated_sentences):
            result.append(translated_sentences[s_idx])
            s_idx += 1
        if b_idx < len(breaks):
            result.append(breaks[b_idx])
            b_idx += 1
    return ''.join(result)

# Setup code to construct valid arguments
translated_sentences = ["Hello", " ", "world", "!"]
breaks = ["\n", "\n", "\n"]

ARGS = {
    "translated_sentences": translated_sentences,
    "breaks": breaks
}
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
replace_awq_linear Not yet run
unverified (harvested, no recorded run)
eole-nlp/eole/eole/modules/awq_linear.py
code served (permissive licence) · get_code("5e8fa1dfb99981ed")
How it ran

no invocation recorded.

replace_bnb_linear Not yet run
unverified (harvested, no recorded run)
eole-nlp/eole/eole/modules/bnb_linear.py
code served (permissive licence) · get_code("21f09e367ce1ec22")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 7 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
pad_at_dim 3 other papers
3 later
2210.06155 (2022-10)
ERNIE-Layout: Layout Knowledge Enhanced Pre-training for Visually-rich Document Understanding
2603.04805 (2026-03)
Attention's Gravitational Field: A Power-Law Interpretation of Positional Correlation
2605.10537 (2026-05)
Mela: Test-Time Memory Consolidation based on Transformation Hypothesis

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

State-of-the-art generic low-precision training algorithms use a mix of 16-bit and 32-bit precision, creating the folklore that 16-bit hardware compute units alone are not enough to maximize model accuracy. As a result, deep learning accelerators are forced to support both 16-bit and 32-bit floating-point units (FPUs), which is more costly than only using 16-bit FPUs for hardware design. We ask: can we train deep learning models only with 16-bit floating-point units, while still matching the model accuracy attained by 32-bit training? Towards this end, we study 16-bit-FPU training on the widely adopted BFloat16 unit. While these units conventionally use nearest rounding to cast output to 16-bit precision, we show that nearest rounding for model weight updates often cancels small updates, which degrades the convergence and model accuracy. Motivated by this, we study two simple techniques well-established in numerical analysis, stochastic rounding and Kahan summation, to remedy the model accuracy degradation in 16-bit-FPU training. We demonstrate that these two techniques can enable up to 7% absolute validation accuracy gain in 16-bit-FPU training. This leads to 0.1% lower to 0.2% higher validation accuracy compared to 32-bit training across seven deep learning applications.

For agents

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

get_harvested_code_for_paper("2010.06192")
get_code_for_paper("2010.06192")
have("2010.06192")

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

curl https://syntology.ai/api/ran/2010.06192.json

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

[![Syntology run record](https://syntology.ai/api/ran/2010.06192.svg)](https://syntology.ai/paper/2010.06192)

Connect an agent — have() is free.