SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2312.00752 · arXiv.org · 2023

Mamba: Linear-Time Sequence Modeling with Selective State Spaces

Albert Gu, Tri Dao

arXiv · PDF · Open in the Atlas

Code that ran

28 of 62 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (7 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 18 have no contract verdict; 9 failures of our instrument; 34 no recorded run), measured 2026-11-21; 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/2312.00752.json · badge: /api/ran/2312.00752.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
vvvm23/mamba-jax community (classifier label)
classifier basis not recorded on the link
6 ran · 6 no recorded run
thearkaprava/ms-temba community (classifier label)
classifier basis not recorded on the link
6 ran · 5 no recorded run
human9000/nd-mamba2-torch community
Papers with Code archive
4 ran · 3 no recorded run
lab-emi/cleanumamba community
Papers with Code archive
2 ran · 10 no recorded run
anzeyimana/deepkin community
Papers with Code archive
2 ran · 5 no recorded run
weitunglin/pixmamba community
Papers with Code archive
2 ran · 0 no recorded run
nus-hpc-ai-lab/dyvm community
Papers with Code archive
1 ran · 1 no recorded run
zzmtsvv/mamba-interface community
not in the paper's declared repositories
1 ran · 1 no recorded run
weiwang31/icemamba community
Papers with Code archive
1 ran · 1 no recorded run
catalpaaa/demansia community
Papers with Code archive
1 ran · 0 no recorded run
hustvl/vim community
Papers with Code archive
1 ran · 0 no recorded run
havenhq/mamba-chat community (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
AmazingDay1/TAME community
Papers with Code archive
0 ran · 1 no recorded run
haiyang-w/tokenformer community
Papers with Code archive
0 ran · 1 no recorded run
FunctionStatusWhere it lives
BiAttn Ran
ran on a synthesized input · constructed an object; no result was computed
weitunglin/pixmamba/classification/models/pixmamba.py
pointer only (licence: NONE) · get_code("b2b4897f1f260e1e")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, weitunglin/pixmamba 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.

LinearActivation Ran
ran; our contract draft was wrong, not the code
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("76fa2650a64a7d15")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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.

Mamba Ran
ran on a synthesized input · constructed an object; no result was computed
nus-hpc-ai-lab/dyvm/Semantic_Segmentation/mamba-1p1p1/mamba_ssm/modules/mamba_simple.py
pointer only (licence: NONE) · get_code("a99d863241faf517")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, nus-hpc-ai-lab/dyvm 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.

Mamba Ran
ran on a synthesized input · constructed an object; no result was computed
catalpaaa/demansia/custom-mamba/mamba_ssm/modules/mamba_simple.py
pointer only (licence: GPL-3.0) · get_code("66eac000db0774c9")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, catalpaaa/demansia, is not in the permissive set Syntology serves inline. 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.

Mamba Ran
ran on a synthesized input · constructed an object; no result was computed
hustvl/vim/mamba-1p1p1/mamba_ssm/modules/mamba_simple.py
code served (permissive licence) · get_code("2aeed50d3dee6266")
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 an instance of its own class, Mamba: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import math

torch.manual_seed(0)

ARGS = {
    'd_model': 4,
    'd_state': 2,
    'd_conv': 4,
    'expand': 2,
    'dt_rank': 'auto',
    'dt_min': 0.001,
    'dt_max': 0.1,
    'dt_init': 'random',
    'dt_scale': 1.0,
    'dt_init_floor': 0.0001,
    'conv_bias': True,
    'bias': False,
    'use_fast_path': True,
    'layer_idx': 0,
    'device': 'cpu',
    'dtype': torch.float32,
    'bimamba_type': 'none',
    'if_divide_out': False,
    'init_layer_scale': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Mamba — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Mamba 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
MambaBlock Ran
ran on a synthesized input · constructed an object; no result was computed
zzmtsvv/mamba-interface/jax/modules.py
pointer only (licence: NONE) · get_code("a58b40a1cca60e8b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, zzmtsvv/mamba-interface 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.

MambaBlock Ran
ran on a synthesized input
anzeyimana/deepkin/deepkin/models/mamba.py
code served (permissive licence) · get_code("ad10c82da33733f2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 3, dtype family float, shape (2, 3, 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
import torch.nn as nn
import math
from dataclasses import dataclass
from typing import Union

# Minimal reproduction of the necessary parts
def log2(x: float) -> float:
    return torch.log(torch.tensor(x)) / torch.log(torch.tensor(2.0))

def npo2(len: int):
    return 2 ** torch.ceil(log2(float(len)))

@dataclass
class MambaConfig:
    d_model: int
    n_layers: int
    dt_rank: Union[int, str] = 'auto'
    d_state: int = 16
    expand_factor: int = 2
    d_conv: int = 4
    norm_epsilon: float = 1e-05
    dt_min: float = 0.001
    dt_max: float = 0.1
    dt_init: str = 'random'
    dt_scale: float = 1.0
    dt_init_floor: float = 0.0001
    bias: bool = False
    conv_bias: bool = True
    pscan: bool = True

    def __post_init__(self):
        self.d_inner = self.expand_factor * self.d_model
        if self.dt_rank == 'auto':
            self.dt_rank = math.ceil(self.d_model / 16)

class MambaBlock(nn.Module):
    def __init__(self, config: MambaConfig):
        super().__init__()
        self.d_model: int = config.d_model
        self.n_layers: int = config.n_layers
        self.dt_rank: int = config.dt_rank if isinstance(config.dt_rank, int) else math.ceil(config.d_model / 16)
        self.d_state: int = config.d_state
        self.expand_factor: int = config.expand_factor
        self.d_conv: int = config.d_conv
        self.norm_epsilon: float = config.norm_epsilon
        self.dt_min: float = config.dt_min
        self.dt_max: float = config.dt_max
        self.dt_init: str = config.dt_init
        self.dt_scale: float = config.dt_scale
        self.dt_init_floor: float = config.dt_init_floor
        self.bias: bool = config.bias
        self.conv_bias: bool = config.conv_bias
        self.pscan: bool = config.pscan
        self.d_inner: int = config.d_inner
        
        self.in_proj = nn.Linear(config.d_model, 2 * config.d_inner, bias=config.bias)
        self.conv1d = nn.Conv1d(
            in_channels=config.d_inner,
            out_channels=config.d_inner,
            kernel_size=config.d_conv,
            bias=config.conv_bias,
            groups=config.d_inner,
            padding=config.d_conv - 1
        )
        self.x_proj = nn.Linear(config.d_inner, self.dt_rank + 2 * config.d_state, bias=False)
        self.dt_proj = nn.Linear(self.dt_rank, config.d_inner, bias=True)
        
        dt_init_std = self.dt_rank ** (-0.5) * config.dt_scale
        if config.dt_init == 'constant':
            nn.init.constant_(self.dt_proj.weight, dt_init_std)
        elif config.dt_init == 'random':
            nn.init.uniform_(self.dt_proj.weight, -dt_init_std, dt_init_std)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        batch, seq_len, d_model = x.shape
        x_proj = self.in_proj(x)
        return x_proj

INIT_ARGS = {
    'config': MambaConfig(
        d_model=4,
        n_layers=2,
        dt_rank='auto',
        d_state=8,
        expand_factor=2,
        d_conv=4,
        norm_epsilon=1e-5,
        dt_min=0.001,
        dt_max=0.1,
        dt_init='random',
        dt_scale=1.0,
        dt_init_floor=0.0001,
        bias=False,
        conv_bias=True,
        pscan=True
    )
}

ARGS = {
    'x': torch.randn(2, 3, 4)
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
RMSNorm Ran
ran on a synthesized input
human9000/nd-mamba2-torch/torchnssd/nd_mamba2.py
pointer only (licence: NONE) · get_code("ba026d72b685bee1")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, human9000/nd-mamba2-torch 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.

SS2D Ran
ran on a synthesized input · constructed an object; no result was computed
weitunglin/pixmamba/classification/models/pixmamba.py
pointer only (licence: NONE) · get_code("c6eb39247d306bdd")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, weitunglin/pixmamba 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.

SS2D Ran
ran on a synthesized input · constructed an object; no result was computed
weiwang31/icemamba/Model/IceMamba/IceMamba_6_VSSB.py
pointer only (licence: NONE) · get_code("75f8a49558b26e64")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, weiwang31/icemamba 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.

append_nvcc_threads Ran
ran on a synthesized input
thearkaprava/ms-temba/mamba-1p1p1/setup_AS.py
code served (permissive licence) · get_code("db4a88f1431b30bd")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

ARGS = {
    'nvcc_extra_args': []
}
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 float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
build_transform Ran
ran; our contract draft was wrong, not the code
thearkaprava/ms-temba/vim/datasets.py
code served (permissive licence) · get_code("ebeb7b6e20a434cf")
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 Compose. 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 torchvision import transforms
from timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD

torch.manual_seed(0)

class ArgsConfig:
    def __init__(self):
        self.input_size = 224
        self.color_jitter = 0.4
        self.aa = 'rand-m9-mstd0.5-inc1'
        self.train_interpolation = 'bicubic'
        self.reprob = 0.25
        self.remode = 'pixel'
        self.recount = 1
        self.eval_crop_ratio = 0.875

args = ArgsConfig()

ARGS = {
    'is_train': True,
    'args': args
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, timm, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Compose — a Compose output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Compose 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
cast_eqx_layer Ran
ran on a synthesized input
vvvm23/mamba-jax/mamba_jax/modelling/equinox/utils.py
code served (permissive licence) · get_code("1cbbea4b9de815f6")
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 Linear. 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 equinox as eqx
import torch
import jax
import jax.numpy as jnp

torch.manual_seed(0)

# Create a simple equinox layer with weight and bias
# We'll use a Linear layer as it has both weight and bias attributes
layer = eqx.nn.Linear(in_features=4, out_features=3, key=jax.random.PRNGKey(0))

# Choose a target dtype - using jnp.float32
dtype = jnp.float32

ARGS = {
    "layer": layer,
    "dtype": dtype,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
equinox, jax, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Linear — a Linear output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Linear 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
chunk_dataset Ran
ran on a synthesized input
vvvm23/mamba-jax/dataset.py
code served (permissive licence) · get_code("1bbd9644f0c6f236")
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 MockDataset. 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
from unittest.mock import MagicMock

# Create a mock dataset with a map method
class MockDataset:
    def __init__(self):
        self.data = {
            'text': [
                'hello world this is a test string',
                'another example for chunking'
            ]
        }
    
    def map(self, fn, batched=True):
        # Apply the map function and return self
        result = fn(self.data)
        self.data = result
        return self

# Create mock args object
class MockArgs:
    def __init__(self):
        self.sequence_length = 10
        self.dataset_text_field = 'text'

# Create dataset instance
dataset = MockDataset()

# Create args instance
args = MockArgs()

# Define ARGS dict
ARGS = {
    'dataset': dataset,
    'args': args,
    'target_field': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MockDataset — a MockDataset output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MockDataset 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
compute_c_state_diversity_loss_simple Ran
ran on a synthesized input
thearkaprava/ms-temba/vim/models_MSTemba.py
code served (permissive licence) · get_code("8f3c3db8de4abd45")
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 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
import torch.nn.functional as F

torch.manual_seed(0)

# Create a list of C states from different dilation groups
# Each C state is a tensor with shape (batch_size, feature_dim)
c_states_list = [
    torch.randn(2, 4),  # batch_size=2, feature_dim=4
    torch.randn(2, 4),  # batch_size=2, feature_dim=4
    torch.randn(2, 4),  # batch_size=2, feature_dim=4
]

ARGS = {
    "c_states_list": c_states_list,
}
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 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
get_logger Ran
ran; our contract draft was wrong, not the code
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("3fcbb53cc7333590")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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.

get_num_layer_for_vit Ran
ran, honoured the contract we drafted
thearkaprava/ms-temba/seg/mmcv_custom/layer_decay_optimizer_constructor.py
code served (permissive licence) · get_code("24385f6e21ea88d7")
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
import torch.nn as nn

torch.manual_seed(0)

# The function takes two parameters:
# 1. var_name: a string representing a variable name
# 2. num_max_layer: an integer representing the maximum number of layers

# Looking at the function body:
# - var_name is compared against strings and used with string operations like startswith() and split()
# - num_max_layer is used as an integer for arithmetic

# We'll test with a realistic var_name from a Vision Transformer backbone
var_name = "backbone.blocks.5.attn.qkv.weight"
num_max_layer = 12

ARGS = {
    "var_name": var_name,
    "num_max_layer": num_max_layer
}
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
grad_wrapper Ran
ran on a synthesized input
vvvm23/mamba-jax/benchmark.py
code served (permissive licence) · get_code("0a8ec528f858d85d")
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 jax
import jax.numpy as jnp

def grad_wrapper(f):
    f_wrapper = lambda *args, **kwargs: f(*args, **kwargs).mean()
    return jax.grad(f_wrapper)

# Create a simple JAX function that takes arguments and returns an array
def simple_jax_function(x, y):
    return x + y

# Test function that takes positional and keyword arguments
def test_function(arr1, arr2=None):
    if arr2 is None:
        arr2 = jnp.ones_like(arr1)
    return arr1 * arr2

# Create small JAX arrays for testing
x = jnp.array([[1.0, 2.0, 3.0, 4.0],
               [5.0, 6.0, 7.0, 8.0]])

y = jnp.array([[0.1, 0.2, 0.3, 0.4],
               [0.5, 0.6, 0.7, 0.8]])

# Define the function to differentiate
def f_to_differentiate(x):
    return x ** 2 + jnp.sin(x)

ARGS = {
    "f": f_to_differentiate
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — 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
mamba_ssm Ran
ran on a synthesized input
vvvm23/mamba-jax/mamba_jax/kernels/reference.py
code served (permissive licence) · get_code("10cfd3bf57dd2545")
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 ArrayImpl, ndim 2, dtype family float, shape (2, 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 jax
import jax.numpy as jnp

# Set random seed for reproducibility
jax.config.update("jax_platform_name", "cpu")

# Define tensor dimensions
l = 2  # sequence length
d_in = 4  # input dimension
n = 3  # state dimension

# Create input tensors
u = jnp.ones((l, d_in), dtype=jnp.float32)
delta = jnp.ones((l, d_in), dtype=jnp.float32) * 0.1
A = jnp.ones((d_in, n), dtype=jnp.float32)
B = jnp.ones((l, n), dtype=jnp.float32)
C = jnp.ones((l, n), dtype=jnp.float32)
D = jnp.ones((d_in,), dtype=jnp.float32)

ARGS = {
    "u": u,
    "delta": delta,
    "A": A,
    "B": B,
    "C": C,
    "D": D,
    "delta_bias": None,
    "delta_softplus": False,
    "associative_scan": True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
new_data_aug_generator Ran
ran; our contract draft was wrong, not the code
thearkaprava/ms-temba/vim/augment.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from aiot-mlsys-lab/famba-v, is pointer only (licence: NONE) · get_code("4a0feeebf8ac43a8")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from aiot-mlsys-lab/famba-v, 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.

preprocess Ran
ran on a synthesized input
havenhq/mamba-chat/trainer/data.py
code served (permissive licence) · get_code("2e14c40a76acb232")
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 transformers
from typing import Dict, Sequence

torch.manual_seed(0)

class StandInTokenizer:
    """Minimal tokenizer stand-in that implements required methods."""
    
    def __init__(self):
        self.use_default_system_prompt = False
    
    def encode(self, text: str, add_special_tokens: bool = True) -> list:
        """Return a list of token IDs."""
        return [1, 2, 3]
    
    def apply_chat_template(self, conversation: Sequence[dict], chat_template: str, max_length: int, truncation: bool) -> list:
        """Apply chat template and return token IDs."""
        # Return a simple token sequence
        return [101, 2054, 2003, 102, 1045, 2572, 2017, 102]

# Create sample conversations
conversations = [
    {
        'messages': [
            {'role': 'user', 'content': 'Hello'},
            {'role': 'assistant', 'content': 'Hi there'},
            {'role': 'user', 'content': 'How are you?'},
            {'role': 'assistant', 'content': 'I am doing well'}
        ]
    },
    {
        'messages': [
            {'role': 'user', 'content': 'What is AI?'},
            {'role': 'assistant', 'content': 'AI is artificial intelligence'}
        ]
    }
]

# Create tokenizer
tokenizer = StandInTokenizer()

# Set parameters
conversation_template = "default"
max_tokens = 128

# Build ARGS dict
ARGS = {
    'conversations': conversations,
    'tokenizer': tokenizer,
    'conversation_template': conversation_template,
    'max_tokens': max_tokens
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, tqdm, transformers — 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
pscan_forward Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
anzeyimana/deepkin/deepkin/models/mamba.py
code served (permissive licence) · get_code("3ccd019afab6e874")
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 Tensor, ndim 4, 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
import torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors with distinct dimensions
# A_in: (B, L, D, N) - batch=2, L=8 (power of 2), D=4, N=3
B = 2
L = 8
D = 4
N = 3

A_in = torch.randn(B, L, D, N)
X_in = torch.randn(B, L, D, N)

ARGS = {
    "A_in": A_in,
    "X_in": X_in,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 — no output shape is stated: observe_signature.py could not observe it (import:OSError)
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
pt_to_raw_pytree Ran
ran on a synthesized input
vvvm23/mamba-jax/mamba_jax/modelling/equinox/loader.py
code served (permissive licence) · get_code("ea69e8168673d498")
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 jax.numpy as jnp

torch.manual_seed(0)

# Create a sample state dict with 'backbone' keys that will be renamed to 'model'
sd = {
    'backbone.layer1.weight': torch.randn(4, 3),
    'backbone.layer1.bias': torch.randn(4),
    'backbone.layer2.weight': torch.randn(5, 4),
    'model.other.weight': torch.randn(2, 3),
}

ARGS = {
    'sd': sd,
    'dtype': jnp.float32,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, 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
resize Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
thearkaprava/ms-temba/vim/models_MSTemba.py
code served (permissive licence) · get_code("55bc02c6846634c4")
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 Tensor, ndim 4, dtype family float, shape (2, 3, 8, 10) 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 input tensor for resizing
# Shape: (batch=2, channels=3, height=4, width=5)
input_tensor = torch.randn(2, 3, 4, 5)

# Define ARGS dict with all parameters
ARGS = {
    'input': input_tensor,
    'size': (8, 10),
    'scale_factor': None,
    'mode': 'nearest',
    'align_corners': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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, 8, 10) — shape (2, 3, 8, 10) 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
segsum Ran
ran; our contract draft was wrong, not the code
human9000/nd-mamba2-torch/torchnssd/nd_mamba2.py
pointer only (licence: NONE) · get_code("d5e983b865a3413d")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, human9000/nd-mamba2-torch 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.

silu Ran
ran; our contract draft was wrong, not the code
human9000/nd-mamba2-torch/torchnssd/nd_mamba2.py
pointer only (licence: NONE) · get_code("8558535d4b8ed416")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, human9000/nd-mamba2-torch 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.

ssd Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
human9000/nd-mamba2-torch/torchnssd/nd_mamba2.py
pointer only (licence: NONE) · get_code("d81ce48746793832")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, human9000/nd-mamba2-torch 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.

wandb_init Ran
ran on a synthesized input
vvvm23/mamba-jax/train_utils.py
code served (permissive licence) · get_code("000e6765f1790cbf")
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 NoopRun. 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


class Args:
    def __init__(self):
        self.wandb = False
        self.learning_rate = 0.001
        self.batch_size = 32
        self.epochs = 10
        self.model_name = "mamba"


ARGS = {
    "args": Args()
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, wandb — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type NoopRun — a NoopRun output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a NoopRun 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
FFTConv Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("e62f17401cf26b98")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

InferenceCache Not yet run
unverified (harvested, no recorded run)
human9000/nd-mamba2-torch/torchnssd/nd_mamba2.py
pointer only (licence: NONE) · get_code("0be091c646cafedd")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, human9000/nd-mamba2-torch 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. No invocation recorded.

Kernel Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("6e0d3f18a6d838b9")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

Mamba Not yet run
unverified (harvested, no recorded run)
AmazingDay1/TAME/mamba_ssm/modules/mamba_simple.py
code served (permissive licence) · get_code("01433a18b73a69d8")
How it ran

no invocation recorded.

Mamba2 Not yet run
unverified (harvested, no recorded run)
human9000/nd-mamba2-torch/torchnssd/nd_mamba2.py
pointer only (licence: NONE) · get_code("b8656f4ec2336e2c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, human9000/nd-mamba2-torch 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. No invocation recorded.

Mamba2Config Not yet run
unverified (harvested, no recorded run)
human9000/nd-mamba2-torch/torchnssd/nd_mamba2.py
pointer only (licence: NONE) · get_code("cba45ceb264445f9")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, human9000/nd-mamba2-torch 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. No invocation recorded.

MambaConfig Not yet run
unverified (harvested, no recorded run)
anzeyimana/deepkin/deepkin/models/mamba.py
code served (permissive licence) · get_code("47df96dc59196907")
How it ran

no invocation recorded.

MambaS4 Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("b6e3bc590d74b59c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

ParallelMambaBlock Not yet run
unverified (harvested, no recorded run)
haiyang-w/tokenformer/megatron/model/mamba/mamba.py
code served (permissive licence) · get_code("b853ef0ddbe24ac6")
How it ran

no invocation recorded.

S4Block Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("58b5b2e41748edc1")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

SSMKernel Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("b8fab631b07b9ceb")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

SSMKernelDPLR Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("cced358b4d4bb38d")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

SSMKernelDiag Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("d5649c5229d3c6f2")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

associative_operator Not yet run
unverified (harvested, no recorded run)
zzmtsvv/mamba-interface/jax/modules.py
pointer only (licence: NONE) · get_code("92d457e0ce6c4038")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, zzmtsvv/mamba-interface 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. No invocation recorded.

build_dataset Not yet run
unverified (harvested, no recorded run)
thearkaprava/ms-temba/vim/datasets.py
code served (permissive licence) · get_code("d58967bba23da4e6")
How it ran

no invocation recorded.

combination Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("2e5e1b8467df7fb2")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

cross_selective_scan Not yet run
unverified (harvested, no recorded run)
weiwang31/icemamba/Model/IceMamba/IceMamba_6_VSSB.py
pointer only (licence: NONE) · get_code("6367281f3513cb50")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, weiwang31/icemamba 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. No invocation recorded.

dplr Not yet run
unverified (harvested, no recorded run)
lab-emi/cleanumamba/src/network/S4/MambaS4.py
pointer only (licence: NONE) · get_code("0e3d9394e01f8e81")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, lab-emi/cleanumamba 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. No invocation recorded.

eval_model Not yet run
unverified (harvested, no recorded run)
thearkaprava/ms-temba/vim/MSTemba_main.py
code served (permissive licence) · get_code("ad88c39d5b497053")
How it ran

no invocation recorded.

get_cuda_bare_metal_version Not yet run
unverified (harvested, no recorded run)
thearkaprava/ms-temba/mamba-1p1p1/setup_AS.py
code served (permissive licence) · get_code("faaf3d02abfa732c")
How it ran

no invocation recorded.

get_pt_checkpoint Not yet run
unverified (harvested, no recorded run)
vvvm23/mamba-jax/mamba_jax/modelling/equinox/loader.py
code served (permissive licence) · get_code("546a5f60b7dbd8fe")
How it ran

no invocation recorded.

init_mamba_from_raw_pytree Not yet run
unverified (harvested, no recorded run)
vvvm23/mamba-jax/mamba_jax/modelling/equinox/loader.py
code served (permissive licence) · get_code("37fea50236b0973b")
How it ran

no invocation recorded.

log2 Not yet run
unverified (harvested, no recorded run)
anzeyimana/deepkin/deepkin/models/mamba.py
code served (permissive licence) · get_code("bf2818a19b4d3706")
How it ran

no invocation recorded.

make_dataset Not yet run
unverified (harvested, no recorded run)
thearkaprava/ms-temba/vim/charades_dataloader.py
code served (permissive licence) · get_code("5984ddd05552649a")
How it ran

no invocation recorded.

make_experiment_directory Not yet run
unverified (harvested, no recorded run)
vvvm23/mamba-jax/train_utils.py
code served (permissive licence) · get_code("be231a069d3f7c48")
How it ran

no invocation recorded.

mamba_inner_with_mask Not yet run
unverified (harvested, no recorded run)
nus-hpc-ai-lab/dyvm/Semantic_Segmentation/mamba-1p1p1/mamba_ssm/modules/mamba_simple.py
pointer only (licence: NONE) · get_code("19e594df117ee0c3")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, nus-hpc-ai-lab/dyvm 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. No invocation recorded.

mamba_ssm Not yet run
unverified (harvested, no recorded run)
vvvm23/mamba-jax/mamba_jax/kernels/interface.py
code served (permissive licence) · get_code("a6bd393ea522c000")
How it ran

no invocation recorded.

npo2 Not yet run
unverified (harvested, no recorded run)
anzeyimana/deepkin/deepkin/models/mamba.py
code served (permissive licence) · get_code("da2a26ac4e6fac71")
How it ran

no invocation recorded.

pad_npo2 Not yet run
unverified (harvested, no recorded run)
anzeyimana/deepkin/deepkin/models/mamba.py
code served (permissive licence) · get_code("1eed9cc6741a3c54")
How it ran

no invocation recorded.

pretokenize_dataset Not yet run
unverified (harvested, no recorded run)
vvvm23/mamba-jax/dataset.py
code served (permissive licence) · get_code("8227c505af8b8afd")
How it ran

no invocation recorded.

pscan Not yet run
unverified (harvested, no recorded run)
anzeyimana/deepkin/deepkin/models/mamba.py
code served (permissive licence) · get_code("98d508bb5b28dd76")
How it ran

no invocation recorded.

run_network Not yet run
unverified (harvested, no recorded run)
thearkaprava/ms-temba/vim/MSTemba_main.py
code served (permissive licence) · get_code("5d474c2e4ca27948")
How it ran

no invocation recorded.

Showing the first 60. An agent gets every one from get_harvested_code_for_paper("2312.00752").

Also present in other papers' code

Of 62 samples lifted from this paper's linked repositories, 62 carry a function identity and 0 do not. Of the 62 distinct functions with an identity, 15 are also present in other papers whose code we hold (11 with at least one paper dated earlier, 1 the same month, 12 later) and 47 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
build_dataset 25 other papers
14 dated earlier, 7 later, 4 same year, order not determinable
2104.05707 (2021-04)
LocalViT: Analyzing Locality in Vision Transformers
2106.00515 (2021-06)
KVT: k-NN Attention for Boosting Vision Transformers
2106.01401 (2021-06)
Container: Context Aggregation Network
and 22 more not named here
build_transform 6 other papers
1 dated earlier, 5 later
2310.01812 (2023-10)
PPT: Token Pruning and Pooling for Efficient Vision Transformers
2403.01590 (2024-03)
The Hidden Attention of Mamba Models
2404.00974 (2024-04)
Improving Visual Recognition with Hyperbolical Visual Hierarchy Mapping
and 3 more not named here
get_num_layer_for_vit 5 other papers
2 dated earlier, 1 the same month, 1 later, 1 same year, order not determinable
2007.09384 (2020-07)
Multi-Scale Positive Sample Refinement for Few-Shot Object Detection
2212.04246 (2022-12)
ViTPose++: Vision Transformer for Generic Body Pose Estimation
and 2 more not named here
new_data_aug_generator 4 other papers
4 later
2403.01590 (2024-03)
The Hidden Attention of Mamba Models
2409.09808 (2024-09)
Famba-V: Fast Vision Mamba with Cross-Layer Token Fusion
2411.04168 (2024-11)
DiMSUM: Diffusion Mamba - A Scalable and Unified Spatial-Frequency Method for Image Generation
and 1 more not named here
Kernel
brick of this paper's assembly: model
2 other papers
1 dated earlier, 1 later
2111.00396 (2021-11)
Efficiently Modeling Long Sequences with Structured State Spaces
2407.08751 (2024-07)
Latent Diffusion for Neural Spiking Data
SSMKernel
brick of this paper's assembly: helper
2 other papers
1 dated earlier, 1 later
2111.00396 (2021-11)
Efficiently Modeling Long Sequences with Structured State Spaces
2407.08751 (2024-07)
Latent Diffusion for Neural Spiking Data
combination
brick of this paper's assembly: helper
2 other papers
1 dated earlier, 1 later
2111.00396 (2021-11)
Efficiently Modeling Long Sequences with Structured State Spaces
2407.08751 (2024-07)
Latent Diffusion for Neural Spiking Data
cross_selective_scan
brick of this paper's assembly: model
2 other papers
2 later
2401.10166 (2024-01)
VMamba: Visual State Space Model
2403.09977 (2024-03)
EfficientVMamba: Atrous Selective Scan for Light Weight Visual Mamba
dplr
brick of this paper's assembly: helper
2 other papers
1 dated earlier, 1 later
2111.00396 (2021-11)
Efficiently Modeling Long Sequences with Structured State Spaces
2407.08751 (2024-07)
Latent Diffusion for Neural Spiking Data
get_cuda_bare_metal_version 2 other papers
2 later
2410.19258 (2024-10)
Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning
2505.23416 (2025-05)
KVzip: Query-Agnostic KV Cache Compression with Context Reconstruction
resize 2 other papers
1 dated earlier, 1 same year, order not determinable
2204.02574 (2022-04)
FocalClick: Towards Practical Interactive Image Segmentation
silu
brick of this paper's assembly: helper
2 other papers
2 later
2405.21060 (2024-05)
Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality
2505.24244 (2025-05)
Mamba Knockout for Unraveling Factual Information Flow
ssm
brick of this paper's assembly: model
2 other papers
1 dated earlier, 1 later
2111.00396 (2021-11)
Efficiently Modeling Long Sequences with Structured State Spaces
2407.08751 (2024-07)
Latent Diffusion for Neural Spiking Data
SSMKernelDPLR
brick of this paper's assembly: model
1 other paper
1 dated earlier
2111.00396 (2021-11)
Efficiently Modeling Long Sequences with Structured State Spaces
SSMKernelDiag
brick of this paper's assembly: model
1 other paper
1 dated earlier
2111.00396 (2021-11)
Efficiently Modeling Long Sequences with Structured State Spaces

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 2026-11-21.

Repositories linked to this paper

More are linked than shown here.

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5× higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation. * Alphabetical by first name.

For agents

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

get_harvested_code_for_paper("2312.00752")
get_code_for_paper("2312.00752")
have("2312.00752")

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

curl https://syntology.ai/api/ran/2312.00752.json

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

[![Syntology run record](https://syntology.ai/api/ran/2312.00752.svg)](https://syntology.ai/paper/2312.00752)

Connect an agent — have() is free.