SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2110.13601 · NeurIPS · 2021

DAG Card is the new Model Card

Jacopo Tagliabue, Ciro Greco, Valay Dave

arXiv · PDF · Open in the Atlas

Code that ran

18 of 25 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 17 have no contract verdict; 1 failure of our instrument; 7 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/2110.13601.json · badge: /api/ran/2110.13601.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
jacopotagliabue/metaflow-intent-prediction canonical (classifier label)
classifier basis not recorded on the link
10 ran · 3 no recorded run
Netflix/metaflow canonical (classifier label)
classifier basis not recorded on the link
1 ran · 2 no recorded run
jacopotagliabue/dag-card-is-the-new-model-card canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
jacopotagliabue/recs-at-resonable-scale community (classifier label)
classifier basis not recorded on the link
6 ran · 2 no recorded run
FunctionStatusWhere it lives
build_html_image_tag Ran
ran on a synthesized input
jacopotagliabue/recs-at-resonable-scale/src/app/pred_inspector.py
code served (permissive licence) · get_code("a8b6d8821b8353f4")
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

ARGS = {
    "image_url": "https://example.com/image.png"
}
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
enable_decorator Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/custom_decorators.py
code served (permissive licence) · get_code("11789a38a836540b")
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 function. 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
from functools import wraps

torch.manual_seed(0)

# Create a simple decorator that wraps a function
def sample_decorator(func):
    @wraps(func)
    def wrapper(*args, **kwargs):
        return func(*args, **kwargs)
    return wrapper

# Create a simple function to decorate
def sample_function(x):
    return x + 1

# Test with flag = "1" (should apply decorator)
# Test with flag = "0" (should not apply decorator)

ARGS = {
    "dec": sample_decorator,
    "flag": "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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
enable_decorator Ran
ran on a synthesized input
jacopotagliabue/recs-at-resonable-scale/src/custom_decorators.py
code served (permissive licence) · get_code("767874914042e842")
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 function. 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


# Create a simple decorator that takes a function and modifies it
def simple_decorator(func):
    def wrapper(*args, **kwargs):
        return func(*args, **kwargs)
    return wrapper


# Create a simple function to be decorated
def example_function(x):
    return x * 2


# Set up arguments for enable_decorator
ARGS = {
    'dec': simple_decorator,
    'flag': '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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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_user_from_tags Ran
ran; our contract draft was wrong, not the code
jacopotagliabue/dag-card-is-the-new-model-card/src/card_builder.py
code served (permissive licence) · get_code("1de5cb6d09d27896")
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) 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.

# Setup code for find_user_from_tags

ARGS = {
    'run_tags': frozenset(['user:alice', 'project:ml-experiment', 'env:prod'])
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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_filename Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/utils.py
code served (permissive licence) · get_code("5513291663ef8a28")
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 get_filename(path):
    return os.path.splitext(os.path.basename(path))[0]

# The function takes a single parameter 'path' which is a file path string
# It uses os.path.basename() to get the filename, then os.path.splitext() to remove the extension
# We need to provide a valid file path string

ARGS = {
    "path": "/home/user/documents/example_file.txt"
}
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
load_module_from_path Ran
ran on a synthesized input
Netflix/metaflow/metaflow/R.py
code served (permissive licence) · get_code("707765fb626eb769")
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 module. 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 tempfile
import os

# Create a temporary Python file to load
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_module.py")

with open(temp_file, "w") as f:
    f.write("""
# Test module for loading
def hello():
    return "Hello from loaded module"

TEST_VAR = 42
""")

ARGS = {
    "module_name": "loaded_test_module",
    "path": temp_file
}
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 module — a module output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a module 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
magicdir Ran
ran on a synthesized input
jacopotagliabue/recs-at-resonable-scale/src/custom_decorators.py
code served (permissive licence) · get_code("d3977dddf3a7e20b")
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 function. 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 shutil
from functools import wraps

# Setup: create a minimal class that the decorator will be applied to
class MockModel:
    def __init__(self):
        self.magicdir = None
    
    def setup_method(self):
        """Method that will be decorated"""
        # Create the magic directory if it doesn't exist
        if not os.path.exists('merlin'):
            os.makedirs('merlin')
        # Create a simple file in it
        with open('merlin/test.txt', 'w') as f:
            f.write('test content')

# Create the decorated function
@wraps(MockModel.setup_method)
def decorated_setup(self):
    """This is what gets called after decoration"""
    from io import BytesIO
    from tarfile import TarFile
    my_magic_dir = 'merlin'
    artifact = 'magicdir'
    
    existing = getattr(self, artifact, None)
    if existing:
        buf = BytesIO(existing)
        with TarFile(mode='r', fileobj=buf) as tar:
            tar.extractall()
    
    # Call original
    if not os.path.exists('merlin'):
        os.makedirs('merlin')
    with open('merlin/test.txt', 'w') as f:
        f.write('test content')
    
    buf = BytesIO()
    with TarFile(mode='w', fileobj=buf) as tar:
        tar.add(my_magic_dir)
    setattr(self, artifact, buf.getvalue())

# Cleanup after
if os.path.exists('merlin'):
    shutil.rmtree('merlin')

# The decorator function takes one parameter: f (a function)
# We pass a simple function as the argument
def simple_func(self):
    pass

ARGS = {
    'f': simple_func
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
make_predictions Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/remote_flow/metaflow/model.py
code served (permissive licence) · get_code("9011d0930fb34a77")
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 ndarray, ndim 2, dtype family float, shape (1, 3) 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 numpy as np
import tensorflow as tf


class MockPredictor:
    """Minimal predictor that implements the predict method."""
    def predict(self, test_inp):
        # test_inp is a dict with 'instances' key containing one-hot encoded data
        instances = test_inp['instances']
        # Return dummy predictions with same batch size
        batch_size = instances.shape[0]
        return np.random.randn(batch_size, 3)


predictor = MockPredictor()

ARGS = {
    'predictor': predictor
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (1, 3) — shape (1, 3) 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
pip Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/custom_decorators.py
code served (permissive licence) · get_code("a3ce5ce1f92ca929")
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 function. 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

torch.manual_seed(0)

# The function 'pip' takes a single parameter 'libraries'
# which is a dict mapping library names (strings) to versions (strings)
# The decorator uses libraries.items() to iterate over the dict

libraries = {
    'numpy': '1.21.0',
    'scipy': '',
    'requests': '2.28.0',
}

ARGS = {
    'libraries': libraries
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
pip Ran
ran on a synthesized input
jacopotagliabue/recs-at-resonable-scale/src/custom_decorators.py
code served (permissive licence) · get_code("df6a316eb33eeb7a")
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 function. 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

torch.manual_seed(0)

# The `pip` function takes a single parameter: `libraries`
# which should be a dict mapping library names (strings) to versions (strings)
# The function uses this dict in a for loop: for library, version in libraries.items()
# and checks if version != ''

ARGS = {
    'libraries': {
        'numpy': '1.21.0',
        'scipy': '',
        'torch': '1.9.0'
    }
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
prepare_dataset Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/prepare_dataset.py
code served (permissive licence) · get_code("ec5dcfa1a6b0f7af")
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 dict. 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 pandas as pd
import tempfile
import os

# Create a temporary parquet file with the required columns
def create_test_parquet():
    # Create sample data with the required columns
    data = {
        'session_id_hash': [1, 1, 1, 2, 2, 2, 3, 3, 3, 3],
        'event_type': ['pageview', 'pageview', 'pageview', 'pageview', 'pageview', 'pageview', 
                       'pageview', 'pageview', 'pageview', 'pageview'],
        'product_action': ['', 'add', '', '', 'add', 'purchase', '', 'add', 'detail', 'purchase']
    }
    df = pd.DataFrame(data)
    
    # Write to a temporary parquet file
    temp_file = tempfile.NamedTemporaryFile(suffix='.parquet', delete=False)
    temp_file.close()
    df.to_parquet(temp_file.name)
    return temp_file.name

# Create the test parquet file
test_parquet_path = create_test_parquet()

ARGS = {
    'training_file': test_parquet_path,
    'K': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
read_sessions_from_training_file Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/prepare_dataset.py
code served (permissive licence) · get_code("07d49ac126d9b329")
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.

import pandas as pd
import tempfile
import os

# Create a temporary parquet file with the required structure
temp_dir = tempfile.mkdtemp()
parquet_file = os.path.join(temp_dir, "test_sessions.parquet")

# Create sample data with the required columns
data = {
    'session_id_hash': ['session_1', 'session_1', 'session_1', 'session_2', 'session_2', 'session_3'],
    'event_type': ['pageview', 'pageview', 'click', 'pageview', 'click', 'pageview'],
    'product_action': ['', '', 'add_to_cart', '', 'purchase', '']
}

df = pd.DataFrame(data)
df.to_parquet(parquet_file)

ARGS = {
    'training_file': parquet_file,
    'K': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — 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
read_to_dataframe Ran
ran on a synthesized input
jacopotagliabue/recs-at-resonable-scale/src/workflow_builder.py
code served (permissive licence) · get_code("da3eeaeed1f008b5")
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 DataFrame. 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 pyarrow as pa
import pandas as pd

# Create a small PyArrow Table
data = {
    'col1': [1, 2, 3],
    'col2': ['a', 'b', 'c'],
    'col3': [1.5, 2.5, 3.5]
}
dataset = pa.table(data)

ARGS = {
    'dataset': dataset,
    'label': 'test_output'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas, pyarrow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
return_json_file_content Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/utils.py
code served (permissive licence) · get_code("27a270038cf86be3")
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 dict. 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 tempfile
import os

# Create a temporary JSON file with sample data
temp_dir = tempfile.mkdtemp()
test_file = os.path.join(temp_dir, "test_data.json")

sample_data = {
    "name": "test",
    "values": [1, 2, 3, 4, 5],
    "nested": {
        "key1": "value1",
        "key2": 42
    }
}

with open(test_file, 'w') as f:
    json.dump(sample_data, f)

ARGS = {
    "file_name": test_file
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
session_indexed Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/remote_flow/metaflow/model.py
code served (permissive licence) · get_code("8f067378d424cbda")
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

torch.manual_seed(0)

# The function takes a session (list of actions) as input
# and converts each action to its corresponding index, adding start/end tokens

ARGS = {
    's': ['add', 'detail', 'view', 'remove', 'purchase']
}
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 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
tf_model_to_tar Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/deploy_model.py
code served (permissive licence) · get_code("b8bb11c469473bde")
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 tempfile
import os
import shutil

# Mock TensorFlow model class since we can't use actual TensorFlow
class MockTFModel:
    def __init__(self):
        self.saved = False
    
    def save(self, filepath):
        """Mock save method that creates the directory structure"""
        os.makedirs(filepath, exist_ok=True)
        # Create a dummy file to represent saved model
        with open(os.path.join(filepath, 'saved_model.pb'), 'w') as f:
            f.write('mock model content')
        self.saved = True

# Set up temporary directory for the test
original_cwd = os.getcwd()
temp_dir = tempfile.mkdtemp()
os.chdir(temp_dir)

try:
    # Create a mock TensorFlow model
    mock_tf_model = MockTFModel()
    
    # Define arguments
    ARGS = {
        'tf_model': mock_tf_model,
        'run_id': 42
    }
finally:
    # Clean up and restore original directory
    os.chdir(original_cwd)
    if os.path.exists(temp_dir):
        shutil.rmtree(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 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
train_lstm_model Ran
ran on a synthesized input
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/model.py
code served (permissive licence) · get_code("2f93893e818afaa6")
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 numpy as np
import tensorflow as tf
import tensorflow.keras as keras

# Create small sample data
np.random.seed(0)

# x: list of session sequences (variable-length lists of integers 0-6)
x = [
    [1, 2, 3],
    [0, 1],
    [2, 3, 4, 5],
    [1, 2],
]

# y: binary labels (purchase or abandon)
y = [1, 0, 1, 0]

# tracker_callback: a simple stand-in callback that implements the Keras Callback interface
class DummyCallback(keras.callbacks.Callback):
    def on_epoch_end(self, epoch, logs=None):
        pass

tracker_callback = DummyCallback()

ARGS = {
    'x': x,
    'y': y,
    'epochs': 2,
    'patience': 1,
    'lstm_dim': 8,
    'batch_size': 2,
    'lr': 0.001,
    'tracker_callback': tracker_callback,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, tensorflow — 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
wrap_response Ran
ran on a synthesized input
jacopotagliabue/recs-at-resonable-scale/src/serverless/handler.py
code served (permissive licence) · get_code("7c153603a40a3b84")
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 dict. 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 time
import uuid

# Examine the function: wrap_response takes status_code (int) and items (list)
# The function uses:
# - status_code: passed to dict
# - items: passed to dict as 'data' value
# - uuid.uuid4(): generates event ID
# - time.time(): generates timestamp
# - json.dumps(): serializes the response
# No special objects or models needed - just basic types

ARGS = {
    'status_code': 200,
    'items': [
        {'id': 1, 'name': 'item1', 'value': 42},
        {'id': 2, 'name': 'item2', 'value': 84},
        {'id': 3, 'name': 'item3', 'value': 126}
    ]
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
encode_image Not yet run
unverified (harvested, no recorded run)
jacopotagliabue/recs-at-resonable-scale/src/app_utils.py
code served (permissive licence) · get_code("9e8d4792ee55f8f8")
How it ran

no invocation recorded.

flow_decorators Not yet run
unverified (harvested, no recorded run)
Netflix/metaflow/metaflow/decorators.py
code served (permissive licence) · get_code("5a0fee822b965644")
How it ran

no invocation recorded.

get_r_func Not yet run
unverified (harvested, no recorded run)
Netflix/metaflow/metaflow/R.py
code served (permissive licence) · get_code("4910597103484b3f")
How it ran

no invocation recorded.

load_clip Not yet run
unverified (harvested, no recorded run)
jacopotagliabue/recs-at-resonable-scale/src/app/pred_inspector.py
code served (permissive licence) · get_code("dd46c001cce00696")
How it ran

no invocation recorded.

process_raw_data Not yet run
unverified (harvested, no recorded run)
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/process_raw_data.py
code served (permissive licence) · get_code("1ba74f174fd45278")
How it ran

no invocation recorded.

read_from_parquet Not yet run
unverified (harvested, no recorded run)
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/process_raw_data.py
code served (permissive licence) · get_code("2c7fed7654dbad7e")
How it ran

no invocation recorded.

return_df Not yet run
unverified (harvested, no recorded run)
jacopotagliabue/metaflow-intent-prediction/local_flow/intent/src/process_raw_data.py
code served (permissive licence) · get_code("17172882ad3ab9e8")
How it ran

no invocation recorded.

Also present in other papers' code

Of 25 samples lifted from this paper's linked repositories, 25 carry a function identity and 0 do not. Of the 25 distinct functions with an identity, 14 are also present in other papers whose code we hold (14 with at least one paper dated earlier, 0 the same month, 0 later) and 11 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
enable_decorator 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
find_user_from_tags 1 other paper
1 dated earlier
1810.03993 (2018-10)
Model Cards for Model Reporting
get_filename 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
make_predictions 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
pip 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
prepare_dataset 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
process_raw_data 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
read_from_parquet 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
read_sessions_from_training_file 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
return_df 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
return_json_file_content 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
session_indexed 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
tf_model_to_tar 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack
train_lstm_model 1 other paper
1 dated earlier
2107.07346 (2021-07)
You Do Not Need a Bigger Boat: Recommendations at Reasonable Scale in a (Mostly) Serverless and Open Stack

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

With the progressive commoditization of modeling capabilities, data-centric AI recognizes that what happens before and after training becomes crucial for realworld deployments. Following the intuition behind Model Cards, we propose DAG Cards as a form of documentation encompassing the tenets of a data-centric point of view. We argue that Machine Learning pipelines (rather than models) are the most appropriate level of documentation for many practical use cases, and we share with the community an open implementation to generate cards from code. * Jacopo designed the first card prototype, which kick-started the project; Ville created Metaflow while working at Netflix and he is one of its main maintainer; Ciro worked on the paper draft and researched the context for the article; finally, Valay implemented cards in Metaflow.

For agents

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

get_harvested_code_for_paper("2110.13601")
get_code_for_paper("2110.13601")
have("2110.13601")

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

curl https://syntology.ai/api/ran/2110.13601.json

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

[![Syntology run record](https://syntology.ai/api/ran/2110.13601.svg)](https://syntology.ai/paper/2110.13601)

Connect an agent — have() is free.