SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.10652 · ACL · 2022

Continual Sequence Generation with Adaptive Compositional Modules

Xuezhi Wang, Diyi Yang, Yanzhe Zhang

arXiv · PDF · Open in the Atlas

Code that ran

14 of 15 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 12 have no contract verdict; 1 failure of our instrument; 1 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/2203.10652.json · badge: /api/ran/2203.10652.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
GT-SALT/Adaptive-Compositional-Modules canonical (classifier label)
classifier basis not recorded on the link
14 ran · 1 no recorded run
FunctionStatusWhere it lives
BN_convert_float Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/fp16util.py
code served (permissive licence) · get_code("e7014fd797aa4e75")
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 SimpleModel. 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)

# Create a module with BatchNorm layers to test the function
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.bn1 = nn.BatchNorm2d(4, affine=True)
        self.bn2 = nn.BatchNorm1d(3, affine=True)
        self.linear = nn.Linear(3, 2)
        self.bn3 = nn.BatchNorm1d(2, affine=True)
    
    def forward(self, x):
        return x

# Instantiate the model
model = SimpleModel()

# Convert model to half precision first so we can test the float conversion
model.half()

ARGS = {
    "module": model
}
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 SimpleModel — a SimpleModel output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleModel 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
conversion_helper Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/fp16.py
code served (permissive licence) · get_code("f887d5ffcbc6ad3b")
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 torch
import torch.nn as nn

torch.manual_seed(0)

def simple_conversion(x):
    """A simple conversion function that doubles a value."""
    if isinstance(x, torch.Tensor):
        return x * 2
    elif isinstance(x, (int, float)):
        return x * 2
    else:
        return x

# Test case 1: Simple scalar value with conversion
val_scalar = 5

# Test case 2: List of scalars
val_list = [1, 2, 3]

# Test case 3: Tuple of scalars
val_tuple = (10, 20, 30)

# Test case 4: Nested list structure
val_nested_list = [1, [2, 3], [[4, 5], 6]]

# Test case 5: Nested tuple structure
val_nested_tuple = (1, (2, 3), ((4, 5), 6))

# Test case 6: Mixed nested structure (list containing tuples)
val_mixed = [1, (2, 3), [4, (5, 6)]]

# Define ARGS with one of the nested structures to demonstrate recursion
ARGS = {
    "val": [1, (2, 3), [[4, 5], 6]],
    "conversion": simple_conversion
}
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 — 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
convert_network Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/fp16util.py
code served (permissive licence) · get_code("6063f8fe87afb580")
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 SimpleNetwork. 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)

# Create a simple network with various module types
class SimpleNetwork(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.bn1 = nn.BatchNorm1d(3, affine=True)
        self.linear2 = nn.Linear(3, 2)
        self.bn2 = nn.BatchNorm1d(2, affine=False)
        self.register_buffer('custom_buffer', torch.randn(2, 3))
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.bn1(x)
        x = self.linear2(x)
        x = self.bn2(x)
        return x

# Instantiate network
network = SimpleNetwork()

# Add a gradient to one parameter to test gradient conversion
network.linear1.weight.grad = torch.randn_like(network.linear1.weight)

# Define target dtype
dtype = torch.float32

ARGS = {
    'network': network,
    'dtype': dtype
}
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 SimpleNetwork — a SimpleNetwork output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a SimpleNetwork 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
convert_pytorch_state_dict_to_flax Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/mytransformers/modeling_flax_pytorch_utils.py
code served (permissive licence) · get_code("b5729b6f3daf1fe2")
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 torch
import numpy as np
from collections import namedtuple

torch.manual_seed(0)

# Create a mock Flax model class
class MockFlaxModel:
    def __init__(self):
        self.base_model_prefix = "model"
        # Create a nested structure of params similar to Flax frozen_dict
        self.params = {
            "model": {
                "layer1": {
                    "kernel": np.zeros((4, 3), dtype=np.float32),
                    "bias": np.zeros((3,), dtype=np.float32),
                },
                "layer2": {
                    "kernel": np.zeros((3, 2), dtype=np.float32),
                    "bias": np.zeros((2,), dtype=np.float32),
                }
            }
        }

# Create PyTorch state dict
pt_state_dict = {
    "model.layer1.weight": torch.randn(3, 4),
    "model.layer1.bias": torch.randn(3),
    "model.layer2.weight": torch.randn(2, 3),
    "model.layer2.bias": torch.randn(2),
}

# Create the mock Flax model
flax_model = MockFlaxModel()

ARGS = {
    "pt_state_dict": pt_state_dict,
    "flax_model": flax_model,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
flax, numpy, torch — 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
convert_tf_weight_name_to_pt_weight_name Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/mytransformers/modeling_tf_pytorch_utils.py
code served (permissive licence) · get_code("5f5d99a0dee8c2b1")
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 re

def convert_tf_weight_name_to_pt_weight_name(tf_name, start_prefix_to_remove=''):
    """
    Convert a TF 2.0 model variable name in a pytorch model weight name.

    Conventions for TF2.0 scopes -> PyTorch attribute names conversions:

        - '$1___$2' is replaced by $2 (can be used to duplicate or remove layers in TF2.0 vs PyTorch)
        - '_._' is replaced by a new level separation (can be used to convert TF2.0 lists in PyTorch nn.ModulesList)

    return tuple with:

        - pytorch model weight name
        - transpose: boolean indicating whether TF2.0 and PyTorch weights matrices are transposed with regards to each
          other
    """
    tf_name = tf_name.replace(':0', '')
    tf_name = re.sub('/[^/]*___([^/]*)/', '/\\1/', tf_name)
    tf_name = tf_name.replace('_._', '/')
    tf_name = re.sub('//+', '/', tf_name)
    tf_name = tf_name.split('/')
    if len(tf_name) > 1:
        tf_name = tf_name[1:]
    transpose = bool(tf_name[-1] in ['kernel', 'pointwise_kernel', 'depthwise_kernel'] or 'emb_projs' in tf_name or 'out_projs' in tf_name)
    if tf_name[-1] == 'kernel' or tf_name[-1] == 'embeddings' or tf_name[-1] == 'gamma':
        tf_name[-1] = 'weight'
    if tf_name[-1] == 'beta':
        tf_name[-1] = 'bias'
    if tf_name[-1] == 'pointwise_kernel' or tf_name[-1] == 'depthwise_kernel':
        tf_name[-1] = tf_name[-1].replace('_kernel', '.weight')
    tf_name = '.'.join(tf_name)
    if start_prefix_to_remove:
        tf_name = tf_name.replace(start_prefix_to_remove, '', 1)
    return (tf_name, transpose)


ARGS = {
    'tf_name': 'layer/dense/kernel:0',
    'start_prefix_to_remove': ''
}
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
fp16_to_fp32 Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/fp16.py
code served (permissive licence) · get_code("38034aa1d784464d")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
from torch.autograd import Variable
from torch.nn.parameter import Parameter

torch.manual_seed(0)

# Test case 1: fp16 tensor
fp16_tensor = torch.randn(2, 4).half()

# Test case 2: fp16 variable
fp16_variable = Variable(torch.randn(2, 4).half())

# Test case 3: fp16 parameter
fp16_parameter = Parameter(torch.randn(2, 4).half())

# Test case 4: nested structure with mixed types
nested_val = (
    fp16_tensor,
    [fp16_variable, fp16_parameter],
    torch.randn(2, 4).half()
)

# Test case 5: fp32 tensor (should pass through unchanged)
fp32_tensor = torch.randn(2, 4)

# Test case 6: mixed fp16 and fp32
mixed_val = [fp16_tensor, fp32_tensor]

ARGS = {
    "val": nested_val
}
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 — 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
fp32_to_fp16 Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/fp16.py
code served (permissive licence) · get_code("ea22baceb9a219a1")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
from torch.nn.parameter import Parameter
from torch.autograd import Variable

torch.manual_seed(0)

# Test with various fp32 tensor types and nested structures
fp32_tensor = torch.randn(2, 4)
fp32_parameter = Parameter(torch.randn(3, 5))
fp32_variable = Variable(torch.randn(2, 4))

# Create nested structures to test the conversion_helper recursion
nested_tuple = (fp32_tensor, fp32_parameter)
nested_list = [fp32_variable, fp32_tensor]
deeply_nested = (fp32_tensor, [fp32_parameter, (fp32_variable, fp32_tensor)])

ARGS = {
    "val": deeply_nested
}
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 — 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_percentile_min_max Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/mytransformers/models/ibert/quant_modules.py
code served (permissive licence) · get_code("613d2566737ae9ec")
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 torch

torch.manual_seed(0)

# Create a small tensor with values that we can test percentile calculations on
# Using shape [10] to have enough elements for meaningful percentile calculations
input_tensor = torch.randn(10)

ARGS = {
    "input": input_tensor,
    "lower_percentile": 0.1,
    "upper_percentile": 99.9,
    "output_tensor": False,
}
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 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
lcs Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/metrics.py
code served (permissive licence) · get_code("0f21a55b40b92463")
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 float, 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 torch

torch.manual_seed(0)

# The function lcs takes two parameters: X and Y
# Looking at the function body:
# - It calls len(X) and len(Y)
# - It indexes into X and Y with [i-1] and [j-1]
# - It compares elements with == (X[i-1] == Y[j-1])
# - It calls lcsstr(X, Y) which also indexes and compares elements
# 
# So X and Y should be sequences (strings, lists, etc.) that support:
# - len()
# - indexing with []
# - element comparison with ==
#
# The simplest choice is to use strings

X = "abcdef"
Y = "fbdamn"

ARGS = {
    "X": X,
    "Y": Y
}
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 float, 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
lcsstr Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/metrics.py
code served (permissive licence) · get_code("989078983378be2e")
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 torch

# lcsstr finds the longest common substring between two strings
# It takes two string parameters and returns an integer
# The function iterates through string1 and string2 comparing characters

string1 = "abcdef"
string2 = "cdefgh"

ARGS = {
    "string1": string1,
    "string2": string2,
}
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 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
linear_quantize Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/mytransformers/models/ibert/quant_modules.py
code served (permissive licence) · get_code("b21b359520c9e1fd")
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 4, dtype family float, shape (2, 3, 4, 4) 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

torch.manual_seed(0)

# Construct arguments for linear_quantize function
# The function quantizes input tensor using scale and zero_point parameters

# Create a 4D input tensor (batch=2, channels=3, height=4, width=4)
input_tensor = torch.randn(2, 3, 4, 4)

# Create scale tensor with shape matching the first dimension of input
scale_tensor = torch.ones(2) * 0.1

# Create zero_point tensor with shape matching the first dimension of input
zero_point_tensor = torch.zeros(2)

# inplace flag
inplace_flag = False

ARGS = {
    "input": input_tensor,
    "scale": scale_tensor,
    "zero_point": zero_point_tensor,
    "inplace": inplace_flag,
}
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 4, dtype family float, shape (2, 3, 4, 4) — shape (2, 3, 4, 4) 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
network_to_half Ran
ran on a synthesized input
GT-SALT/Adaptive-Compositional-Modules/fp16util.py
code served (permissive licence) · get_code("9919cae853c6180f")
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 Sequential. 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)

# Create a simple neural network with BatchNorm layers
# The function will call .half() on it and convert BatchNorms to float
class SimpleNetworkWithBN(nn.Module):
    def __init__(self):
        super(SimpleNetworkWithBN, self).__init__()
        self.conv1 = nn.Conv2d(3, 4, kernel_size=3)
        self.bn1 = nn.BatchNorm2d(4)
        self.linear = nn.Linear(4, 2)
    
    def forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        return x

# Instantiate the network
network = SimpleNetworkWithBN()

# Define ARGS dict
ARGS = {
    "network": network
}
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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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
normalize_text Ran
ran; our contract draft was wrong, not the code
GT-SALT/Adaptive-Compositional-Modules/metrics.py
code served (permissive licence) · get_code("b971ea3b818d09b7")
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 string
import re

def normalize_text(s):
    """Lower text and remove punctuation, articles and extra whitespace."""

    def remove_articles(text):
        return re.sub('\\b(a|an|the)\\b', ' ', text)

    def white_space_fix(text):
        return ' '.join(text.split())

    def remove_punc(text):
        exclude = set(string.punctuation)
        return ''.join((ch for ch in text if ch not in exclude))

    def lower(text):
        return text.lower()
    return white_space_fix(remove_articles(remove_punc(lower(s))))


ARGS = {
    "s": "The Quick Brown Fox! Jumps over a lazy dog."
}
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
to_python_float Ran
ran, honoured the contract we drafted
GT-SALT/Adaptive-Compositional-Modules/loss_scaler.py
pointer only (licence: MIT); the stored copy get_code serves, from lengsicong/tell2design, is pointer only (licence: NONE) · get_code("6b5d363ab4dbf012")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from lengsicong/tell2design, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

symmetric_linear_quantization_params Not yet run
unverified (harvested, no recorded run)
GT-SALT/Adaptive-Compositional-Modules/mytransformers/models/ibert/quant_modules.py
code served (permissive licence) · get_code("fca1d8b44670f871")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 11 are also present in other papers whose code we hold (11 with at least one paper dated earlier, 0 the same month, 11 later) and 4 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
to_python_float 9 other papers
3 dated earlier, 5 later, 1 same year, order not determinable
1904.00962 (2019-04)
Large Batch Optimization for Deep Learning: Training BERT in 76 minutes
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
2109.04332 (2021-09)
PPT: Pre-trained Prompt Tuning for Few-shot Learning
and 6 more not named here
conversion_helper 4 other papers
3 dated earlier, 1 later
1803.09017 (2018-03)
Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis
1812.04342 (2018-12)
Learning latent representations for style control and transfer in end-to-end speech synthesis
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
and 1 more not named here
fp16_to_fp32 4 other papers
3 dated earlier, 1 later
1803.09017 (2018-03)
Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis
1812.04342 (2018-12)
Learning latent representations for style control and transfer in end-to-end speech synthesis
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
and 1 more not named here
fp32_to_fp16 4 other papers
3 dated earlier, 1 later
1803.09017 (2018-03)
Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis
1812.04342 (2018-12)
Learning latent representations for style control and transfer in end-to-end speech synthesis
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
and 1 more not named here
convert_tf_weight_name_to_pt_weight_name 3 other papers
1 dated earlier, 1 later, 1 same year, order not determinable
2109.03808 (2021-09)
Smelting Gold and Silver for Improved Multilingual AMR-to-Text Generation
2303.14582 (2023-03)
Identification of Negative Transfers in Multitask Learning Using Surrogate Models
normalize_text 3 other papers
2 dated earlier, 1 later
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
2006.13009 (2020-06)
Iterative Deep Graph Learning for Graph Neural Networks: Better and Robust Node Embeddings
2025.emnlp-industry.89 (2025 (year only))
L4: Mutual Learning Helps Lifelong Language Learning
BN_convert_float 2 other papers
1 dated earlier, 1 later
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
2025.emnlp-industry.89 (2025 (year only))
L4: Mutual Learning Helps Lifelong Language Learning
convert_network 2 other papers
1 dated earlier, 1 later
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
2025.emnlp-industry.89 (2025 (year only))
L4: Mutual Learning Helps Lifelong Language Learning
lcs 2 other papers
1 dated earlier, 1 later
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
2025.emnlp-industry.89 (2025 (year only))
L4: Mutual Learning Helps Lifelong Language Learning
lcsstr 2 other papers
1 dated earlier, 1 later
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
2025.emnlp-industry.89 (2025 (year only))
L4: Mutual Learning Helps Lifelong Language Learning
network_to_half 2 other papers
1 dated earlier, 1 later
1909.03329 (2019-09)
LAMOL: LAnguage MOdeling for Lifelong Language Learning
2025.emnlp-industry.89 (2025 (year only))
L4: Mutual Learning Helps Lifelong Language Learning

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

Continual learning is essential for real-world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks. Existing work on continual sequence generation either always reuses existing parameters to learn new tasks, which is vulnerable to catastrophic forgetting on dissimilar tasks, or blindly adds new parameters for every new task, which could prevent knowledge sharing between similar tasks. To get the best of both worlds, in this work, we propose continual sequence generation with adaptive compositional modules to adaptively add modules in transformer architectures and compose both old and new modules for new tasks. We also incorporate pseudo experience replay to facilitate knowledge transfer in those shared modules. Experiment results on various sequences of generation tasks show that our framework can adaptively add modules or reuse modules based on task similarity, outperforming state-of-the-art baselines in terms of both performance and parameter efficiency. We make our code public at https://github.com/GT-SALT/ Adaptive-Compositional-Modules.

For agents

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

get_harvested_code_for_paper("2203.10652")
get_code_for_paper("2203.10652")
have("2203.10652")

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

curl https://syntology.ai/api/ran/2203.10652.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.10652.svg)](https://syntology.ai/paper/2203.10652)

Connect an agent — have() is free.