SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2202.00512 · ICLR · 2022

PROGRESSIVE DISTILLATION FOR FAST SAMPLING OF DIFFUSION MODELS

Tim Salimans, Jonathan Ho

arXiv · PDF · Open in the Atlas

Code that ran

10 of 27 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (4 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 6 have no contract verdict; 3 failures of our instrument; 17 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/2202.00512.json · badge: /api/ran/2202.00512.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
sainzerjj/sferd community
Papers with Code archive
2 ran · 3 no recorded run
deepxuan/dn-dp community
Papers with Code archive
2 ran · 2 no recorded run
ericl122333/patchdiffusion-pytorch community
Papers with Code archive
2 ran · 1 no recorded run
google-research/google-research unrelated (classifier label)
classifier basis: classifier_2025
1 ran · 6 no recorded run
Hramchenko/diffusion_distiller community
Papers with Code archive
1 ran · 3 no recorded run
cloneofsimo/consistency_models community
Papers with Code archive
1 ran · 0 no recorded run
sreerajr000/consistency-models community
Papers with Code archive
1 ran · 0 no recorded run
viniciusmikuni/gsgm community
Papers with Code archive
0 ran · 2 no recorded run
FunctionStatusWhere it lives
ConsistencyModel Ran
ran on a synthesized input · constructed an object; no result was computed
cloneofsimo/consistency_models/consistency_models/unet.py
pointer only (licence: NONE) · get_code("c5b3831528f9c8ce")
How it ran

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

EMA Ran
ran on a synthesized input
sainzerjj/sferd/diffusion/train_utils.py
code served (permissive licence) · get_code("1baf4c2b5fd0b943")
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 NoneType. 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 weakref

class EMA:
    """
    exponential moving average
    inspired by:
    [1] https://github.com/fadel/pytorch_ema
    [2] https://github.com/tensorflow/tensorflow/blob/v2.9.1/tensorflow/python/training/moving_averages.py#L281-L685
    """

    def __init__(self, model, decay=0.9999):
        shadow = []
        refs = []
        for k, v in model.named_parameters():
            if v.requires_grad:
                shadow.append((k, v.detach().clone()))
                refs.append((k, weakref.ref(v)))
        self.shadow = dict(shadow)
        self._refs = dict(refs)
        self.decay = decay
        self.num_updates = 0
        self.backup = None

    def update(self):
        self.num_updates += 1
        decay = min(self.decay, (1 + self.num_updates) / (10 + self.num_updates))
        for k, _ref in self._refs.items():
            assert _ref() is not None, 'referenced object no longer exists!'
            self.shadow[k] += (1 - decay) * (_ref().data - self.shadow[k])

    def apply(self):
        self.backup = dict([(k, _ref().detach().clone()) for k, _ref in self._refs.items()])
        for k, _ref in self._refs.items():
            _ref().data.copy_(self.shadow[k])

    def restore(self):
        for k, _ref in self._refs.items():
            _ref().data.copy_(self.backup[k])
        self.backup = None

    def __enter__(self):
        self.apply()

    def __exit__(self, *exc):
        self.restore()

    def state_dict(self):
        return {'decay': self.decay, 'shadow': self.shadow, 'num_updates': self.num_updates}

    @property
    def extra_states(self):
        return {'decay', 'num_updates'}

    def load_state_dict(self, state_dict, strict=True):
        _dict_keys = set(self.__dict__['shadow']).union(self.extra_states)
        dict_keys = set(state_dict['shadow']).union(self.extra_states)
        incompatible_keys = set.symmetric_difference(_dict_keys, dict_keys) if strict else set.difference(_dict_keys, dict_keys)
        if incompatible_keys:
            raise RuntimeError(f"Key mismatch!\nMissing key(s): {', '.join(set.difference(_dict_keys, dict_keys))}.Unexpected key(s): {', '.join(set.difference(dict_keys, _dict_keys))}")
        self.__dict__.update(state_dict)


model = nn.Sequential(
    nn.Linear(4, 8),
    nn.ReLU(),
    nn.Linear(8, 3)
)

INIT_ARGS = {
    'model': model,
    'decay': 0.9999
}

ARGS = {}

METHOD = 'update'

ema = EMA(**INIT_ARGS)
ema.update()
ema.apply()
ema.restore()
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 NoneType — a NoneType output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a NoneType 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
GaussianDiffusion Ran
ran on a synthesized input · constructed an object; no result was computed
ericl122333/patchdiffusion-pytorch/patch_diffusion/respace.py
pointer only (licence: NONE) · get_code("afb6a311f3326182")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, ericl122333/patchdiffusion-pytorch 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.

Model Ran
ran on a synthesized input · constructed an object; no result was computed
google-research/google-research/diffusion_distillation/diffusion_distillation/dpm.py
code served (permissive licence) · get_code("e6c610e22a0b8e1f")
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, Model: 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 jax
import jax.numpy as jnp
import numpy as onp
from flax import linen as nn

# Mock utils module since it's imported in the function
class utils:
    @staticmethod
    def broadcast_from_left(x, shape):
        """Broadcast x to shape from the left."""
        while len(x.shape) < len(shape):
            x = jnp.expand_dims(x, axis=-1)
        return jnp.broadcast_to(x, shape)
    
    @staticmethod
    def log1mexp(x):
        """Numerically stable log(1 - exp(x))."""
        return jnp.log(-jnp.expm1(x))

# Create a simple model function that takes z and logsnr and returns predictions
def simple_model_fn(z, logsnr):
    """Simple model that outputs epsilon predictions."""
    # Output same shape as z for 'eps' mean_type
    return jnp.zeros_like(z)

# Alternative model function for 'both' mean_type
def model_fn_both(z, logsnr):
    """Model that outputs both x and eps predictions."""
    # Output 2x the channel dimension for 'both' mean_type
    batch_size = z.shape[0]
    spatial_dims = z.shape[1:-1]
    channels = z.shape[-1]
    return jnp.zeros((batch_size,) + spatial_dims + (2 * channels,))

# Create minimal Model class instances with different configurations
model_eps = type('Model', (), {
    'model_fn': simple_model_fn,
    'mean_type': 'eps',
    'logvar_type': 'small',
    'logvar_coeff': 0.1,
    'target_model_fn': None,
})()

model_x = type('Model', (), {
    'model_fn': simple_model_fn,
    'mean_type': 'x',
    'logvar_type': 'small',
    'logvar_coeff': 0.1,
    'target_model_fn': None,
})()

model_v = type('Model', (), {
    'model_fn': simple_model_fn,
    'mean_type': 'v',
    'logvar_type': 'large',
    'logvar_coeff': 0.2,
    'target_model_fn': None,
})()

model_both = type('Model', (), {
    'model_fn': model_fn_both,
    'mean_type': 'both',
    'logvar_type': 'medium:0.5',
    'logvar_coeff': 0.15,
    'target_model_fn': None,
})()

# Create small test tensors
batch_size = 2
spatial_dim = 4
channels = 3

z_t = jnp.ones((batch_size, spatial_dim, channels))
logsnr_t = jnp.array([0.5, 1.0])
logsnr_s = jnp.array([0.2, 0.8])
x = jnp.ones((batch_size, spatial_dim, channels))
x_logvar_small = 'small'
x_logvar_large = 'large'
x_logvar_medium = 'medium:0.5'
x_logvar_array = jnp.zeros((batch_size, spatial_dim, channels))

ARGS = {
    'model_fn': simple_model_fn,
    'mean_type': 'eps',
    'logvar_type': 'small',
    'logvar_coeff': 0.1,
    'target_model_fn': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
absl, flax, jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Model — 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 Model 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
RunningStatistics Ran
ran on a synthesized input
sainzerjj/sferd/diffusion/train_utils.py
code served (permissive licence) · get_code("0b2e4925d1642a0b")
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 NoneType. 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

INIT_ARGS = {
    'loss': 0.5,
    'accuracy': 0.8,
    'time': 1.2
}

ARGS = {
    'n': 4,
    'loss': 0.3,
    'accuracy': 0.9,
    'time': 0.5
}

METHOD = 'update'

# Construct and call the class
if __name__ == '__main__':
    from torch import nn
    
    # Instantiate with INIT_ARGS
    runner = RunningStatistics(**INIT_ARGS)
    print("Initial state:")
    print(runner)
    print()
    
    # Call update method with ARGS
    runner.update(**ARGS)
    print("After first update:")
    print(runner)
    print()
    
    # Call extract to get averages
    result = runner.extract()
    print("Extracted averages:")
    print(result)
    print()
    
    # Call reset
    runner.reset()
    print("After reset:")
    print(runner)


class RunningStatistics:

    def __init__(self, **kwargs):
        self.count = 0
        self.stats = []
        for k, v in kwargs.items():
            self.stats.append((k, v or 0))
        self.stats = dict(self.stats)

    def reset(self):
        self.count = 0
        for k in self.stats:
            self.stats[k] = 0

    def update(self, n, **kwargs):
        self.count += n
        for k, v in kwargs.items():
            self.stats[k] = self.stats.get(k, 0) + v

    def extract(self):
        avg_stats = []
        for k, v in self.stats.items():
            avg_stats.append((k, v / self.count))
        return dict(avg_stats)

    def __repr__(self):
        out_str = 'Count(s): {}\n'
        out_str += 'Statistics:\n'
        for k in self.stats:
            out_str += f'\t{k} = {{{k}}}\n'
        return out_str.format(self.count, **self.stats)
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type NoneType — a NoneType output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a NoneType 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
_WrappedModel Ran
ran on a synthesized input · constructed an object; no result was computed
ericl122333/patchdiffusion-pytorch/patch_diffusion/respace.py
pointer only (licence: NONE) · get_code("eeb6b2016ef36e61")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, ericl122333/patchdiffusion-pytorch 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.

_extract_into_tensor Ran
ran; our contract draft was wrong, not the code
deepxuan/dn-dp/model/sr3_modules/respace.py
pointer only (licence: NONE) · get_code("7cf7aa019c5dcc57")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, deepxuan/dn-dp 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.

make_beta_schedule Ran
ran, honoured the contract we drafted
deepxuan/dn-dp/model/sr3_modules/respace.py
pointer only (licence: NONE) · get_code("7ef3b9728402cdd1")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, deepxuan/dn-dp 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.

make_none_args Ran
ran; our contract draft was wrong, not the code
Hramchenko/diffusion_distiller/train_utils.py
code served (permissive licence) · get_code("513b57c8cf771713")
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 torch.nn as nn

torch.manual_seed(0)

# Create a small image tensor (batch=2, channels=3, height=4, width=4)
img = torch.randn(2, 3, 4, 4)

# Create a small label tensor (batch=2)
label = torch.randint(0, 10, (2,))

# Device set to CPU
device = 'cpu'

ARGS = {
    'img': img,
    'label': label,
    'device': device
}
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 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
multistep_consistency_sampling Ran
ran; our contract draft was wrong, not the code
sreerajr000/consistency-models/sampler.py
code served (permissive licence) · get_code("18b1cad6bd142108")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

class SimpleConsistencyNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
        self.sigma_min = 0.002
    
    def forward(self, x, t):
        # x: latents tensor
        # t: timestep (scalar tensor)
        return self.linear(x)

net = SimpleConsistencyNet()
latents = torch.randn(2, 4)
t_steps = [0.1, 0.05, 0.01]

ARGS = {
    "net": net,
    "latents": latents,
    "t_steps": t_steps,
}
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 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
DeepSetsAtt Not yet run
unverified (harvested, no recorded run)
viniciusmikuni/gsgm/demo/GSGM.py
pointer only (licence: NONE) · get_code("9cde3092e14a7d62")
How it ran

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

DiffusionDistillation Not yet run
unverified (harvested, no recorded run)
Hramchenko/diffusion_distiller/train_utils.py
code served (permissive licence) · get_code("791b3bf86655f018")
How it ran

no invocation recorded.

DistillationTrainer Not yet run
unverified (harvested, no recorded run)
sainzerjj/sferd/diffusion/train_utils.py
code served (permissive licence) · get_code("b13671cc03127267")
How it ran

no invocation recorded.

DummyScheduler Not yet run
unverified (harvested, no recorded run)
sainzerjj/sferd/diffusion/train_utils.py
code served (permissive licence) · get_code("d56db955dcd1e1bb")
How it ran

no invocation recorded.

GSGM Not yet run
unverified (harvested, no recorded run)
viniciusmikuni/gsgm/demo/GSGM.py
pointer only (licence: NONE) · get_code("ef2ea1d8599e8b2d")
How it ran

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

GaussianDiffusion Not yet run
unverified (harvested, no recorded run)
deepxuan/dn-dp/model/sr3_modules/respace.py
pointer only (licence: NONE) · get_code("92244ba45f2a5ca4")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, deepxuan/dn-dp 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.

SpacedDiffusion Not yet run
unverified (harvested, no recorded run)
ericl122333/patchdiffusion-pytorch/patch_diffusion/respace.py
pointer only (licence: NONE) · get_code("3ef0f294c952f62d")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, ericl122333/patchdiffusion-pytorch 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.

SpacedDiffusion Not yet run
unverified (harvested, no recorded run)
deepxuan/dn-dp/model/sr3_modules/respace.py
pointer only (licence: NONE) · get_code("0bceb8d35b696f6a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, deepxuan/dn-dp 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.

default_iter_callback Not yet run
unverified (harvested, no recorded run)
Hramchenko/diffusion_distiller/train_utils.py
code served (permissive licence) · get_code("7b619a5570fecf04")
How it ran

no invocation recorded.

diffusion_forward Not yet run
unverified (harvested, no recorded run)
google-research/google-research/diffusion_distillation/diffusion_distillation/dpm.py
code served (permissive licence) · get_code("d37af918412dc794")
How it ran

no invocation recorded.

diffusion_reverse Not yet run
unverified (harvested, no recorded run)
google-research/google-research/diffusion_distillation/diffusion_distillation/dpm.py
code served (permissive licence) · get_code("e6029bbacbfd72ee")
How it ran

no invocation recorded.

moving_average Not yet run
unverified (harvested, no recorded run)
Hramchenko/diffusion_distiller/train_utils.py
code served (permissive licence) · get_code("2dae0e9618d089f2")
How it ran

no invocation recorded.

predict_eps_from_x Not yet run
unverified (harvested, no recorded run)
google-research/google-research/diffusion_distillation/diffusion_distillation/dpm.py
code served (permissive licence) · get_code("ecf3ebad654f83f5")
How it ran

no invocation recorded.

predict_v_from_x_and_eps Not yet run
unverified (harvested, no recorded run)
google-research/google-research/diffusion_distillation/diffusion_distillation/dpm.py
code served (permissive licence) · get_code("894c180fa74bdbd5")
How it ran

no invocation recorded.

predict_x_from_eps Not yet run
unverified (harvested, no recorded run)
google-research/google-research/diffusion_distillation/diffusion_distillation/dpm.py
code served (permissive licence) · get_code("66e7c354db444496")
How it ran

no invocation recorded.

predict_x_from_v Not yet run
unverified (harvested, no recorded run)
google-research/google-research/diffusion_distillation/diffusion_distillation/dpm.py
code served (permissive licence) · get_code("a8cf5ffc562524b9")
How it ran

no invocation recorded.

save_image Not yet run
unverified (harvested, no recorded run)
sainzerjj/sferd/diffusion/train_utils.py
code served (permissive licence) · get_code("d9b8534277a98f88")
How it ran

no invocation recorded.

Also present in other papers' code

Of 27 samples lifted from this paper's linked repositories, 27 carry a function identity and 0 do not. Of the 27 distinct functions with an identity, 4 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 4 later) and 23 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
make_beta_schedule
brick of this paper's assembly: helper
10 other papers
1 dated earlier, 9 later · 3 only as an assembly brick
2104.07636 (2021-04)
Image Super-Resolution via Iterative Refinement
2206.11892 (2022-06)
DDPM-CD: Denoising Diffusion Probabilistic Models as Feature Extractors for Change Detection
2209.14566 (2022-09)
Diffusion Adversarial Representation Learning for Self-supervised Vessel Segmentation
and 7 more not named here
SpacedDiffusion
brick of this paper's assembly: entry
7 other papers
1 dated earlier, 6 later
2010.02502 (2020-10)
Denoising Diffusion Implicit Models
2203.08382 (2022-03)
Dual Diffusion Implicit Bridges for Image-to-Image Translation
2310.11320 (2023-10)
Towards Generic Semi-Supervised Framework for Volumetric Medical Image Segmentation
and 4 more not named here
_WrappedModel
brick of this paper's assembly: helper
7 other papers
1 dated earlier, 6 later · 7 only as an assembly brick
2010.02502 (2020-10)
Denoising Diffusion Implicit Models
2203.08382 (2022-03)
Dual Diffusion Implicit Bridges for Image-to-Image Translation
2306.17775 (2023-06)
Practical and Asymptotically Exact Conditional Sampling in Diffusion Models
and 4 more not named here
GaussianDiffusion
brick of this paper's assembly: helper
1 other paper
1 later
2204.00227 (2022-04)
Perception Prioritized Training of Diffusion Models

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

Diffusion models have recently shown great promise for generative modeling, outperforming GANs on perceptual quality and autoregressive models at density estimation. A remaining downside is their slow sampling time: generating high quality samples takes many hundreds or thousands of model evaluations. Here we make two contributions to help eliminate this downside: First, we present new parameterizations of diffusion models that provide increased stability when using few sampling steps. Second, we present a method to distill a trained deterministic diffusion sampler, using many steps, into a new diffusion model that takes half as many sampling steps. We then keep progressively applying this distillation procedure to our model, halving the number of required sampling steps each time. On standard image generation benchmarks like CIFAR-10, ImageNet, and LSUN, we start out with state-of-the-art samplers taking as many as 8192 steps, and are able to distill down to models taking as few as 4 steps without losing much perceptual quality; achieving, for example, a FID of 3.0 on CIFAR-10 in 4 steps. Finally, we show that the full progressive distillation procedure does not take more time than it takes to train the original model, thus representing an efficient solution for generative modeling using diffusion at both train and test time.

For agents

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

get_harvested_code_for_paper("2202.00512")
get_code_for_paper("2202.00512")
have("2202.00512")

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

curl https://syntology.ai/api/ran/2202.00512.json

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

[![Syntology run record](https://syntology.ai/api/ran/2202.00512.svg)](https://syntology.ai/paper/2202.00512)

Connect an agent — have() is free.