Diffusion API

This module provides discrete and continuous diffusion models for generative modeling.

D3PM (Discrete Diffusion)

class medlatents.diffusion.D3PM(num_classes, num_timesteps=1000, schedule_type='cosine', transition_type='absorbing', hybrid_loss_coeff=0.001, device=None)

Bases: object

Discrete Denoising Diffusion Probabilistic Model.

Supports multiple transition types: - absorbing: Transitions to a mask state (like masked language modeling) - uniform: Uniform transitions to all classes (original D3PM) - gaussian: Discretized Gaussian transitions (smooth blurring)

Parameters:

• num_classes (int)

• num_timesteps (int, default: 1000)

• schedule_type (Literal['linear', 'cosine', 'quadratic', 'sigmoid'], default: 'cosine')

• transition_type (Literal['absorbing', 'uniform', 'gaussian'], default: 'absorbing')

• hybrid_loss_coeff (float, default: 0.001)

• device (device | None, default: None)

__init__(num_classes, num_timesteps=1000, schedule_type='cosine', transition_type='absorbing', hybrid_loss_coeff=0.001, device=None)

Parameters:

• num_classes (int)

• num_timesteps (int, default: 1000)

• schedule_type (Literal['linear', 'cosine', 'quadratic', 'sigmoid'], default: 'cosine')

• transition_type (Literal['absorbing', 'uniform', 'gaussian'], default: 'absorbing')

• hybrid_loss_coeff (float, default: 0.001)

• device (device | None, default: None)

compile_sampling(mode='reduce-overhead', fullgraph=False, dynamic=True)

Compile the sampling step using torch.compile for faster inference.

Requires PyTorch 2.0+. On older versions, this is a no-op.

Parameters:

• mode (str, default: 'reduce-overhead') – Compilation mode. Options: - “default”: Good balance of compile time and runtime - “reduce-overhead”: Minimize overhead, best for small models - “max-autotune”: Maximum optimization, longer compile time

• fullgraph (bool, default: False) – If True, require the entire function to be captured as a single graph

• dynamic (bool, default: True) – If True, enable dynamic shape support (recommended for variable batch/seq)

Return type:

bool

Returns:

True if compilation was successful, False if torch.compile unavailable

q_sample(x_0, t, noise=None)

Sample from q(x_t | x_0) - the forward diffusion process.

Parameters:

• x_0 (Int[Tensor, 'batch seq']) – Clean data [batch, seq_len]

• t (Int[Tensor, 'batch']) – Timesteps [batch]

• noise (Tensor | None, default: None) – Optional pre-sampled noise

Returns:

Noisy data at timestep t

Return type:

x_t

q_posterior(x_0, x_t, t)

Compute the posterior distribution q(x_{t-1} | x_t, x_0).

This computes the true posterior for the reverse process, which is used for training. The posterior is computed using Bayes’ rule:

q(x_{t-1} | x_t, x_0) ∝ q(x_t | x_{t-1}) · q(x_{t-1} | x_0)

Where: - q(x_t | x_{t-1}) comes from the transition matrix Q_t - q(x_{t-1} | x_0) comes from the cumulative transition Q̄_{t-1}

Parameters:

• x_0 (Int[Tensor, 'batch seq']) – Clean data [batch, seq_len]

• x_t (Int[Tensor, 'batch seq']) – Noisy data at timestep t [batch, seq_len]

• t (Int[Tensor, 'batch']) – Timesteps [batch]

Return type:

Float[Tensor, 'batch seq num_classes']

Returns:

Normalized posterior probabilities [batch, seq_len, num_classes]

p_sample(model, x_t, t, temperature=1.0, **model_kwargs)

Sample from p(x_{t-1} | x_t) - the learned reverse process.

Uses vectorized marginalization over all x_0 values simultaneously. If compile_sampling() was called, uses the compiled computation graph.

Parameters:

• model (Module) – Denoising model that predicts p(x_0 | x_t) or p(x_{t-1} | x_t)

• x_t (Int[Tensor, 'batch seq']) – Noisy data at timestep t

• t (Int[Tensor, 'batch']) – Timesteps

• temperature (float, default: 1.0) – Sampling temperature

• **model_kwargs – Additional arguments for model

Returns:

Sample at timestep t-1

Return type:

x_{t-1}

compute_loss(model, x_0, t=None, loss_type='hybrid', **model_kwargs)

Compute training loss.

Parameters:

• model (Module) – Denoising model

• x_0 (Int[Tensor, 'batch seq']) – Clean data

• t (Int[Tensor, 'batch'] | None, default: None) – Timesteps (randomly sampled if None)

• loss_type (Literal['vb', 'hybrid', 'cross_entropy'], default: 'hybrid') – Type of loss to compute

• **model_kwargs – Additional model arguments

Returns:

Scalar loss

Return type:

loss

sample(model, shape, temperature=1.0, **model_kwargs)

Generate samples using the reverse diffusion process.

Parameters:

• model (Module) – Trained denoising model

• shape (tuple) – Shape of samples (batch_size, seq_len)

• temperature (float, default: 1.0) – Sampling temperature

• **model_kwargs – Additional model arguments

Returns:

Generated samples

Return type:

x_0

class medlatents.diffusion.D3PMKLASS(d3pm, kl_threshold=0.01, min_steps=50, window_size=5)

Bases: object

KL-Adaptive Stability Sampling (KLASS) for D3PM/discrete diffusion.

Dynamically stops sampling when predictions stabilize, rather than using all timesteps. Can significantly speed up inference.

Reference: “KL-Guided Fast Inference in Masked Diffusion Models” (arXiv:2511.05664)

Parameters:

• d3pm (D3PM) – Base D3PM instance

• kl_threshold (float, default: 0.01) – Stop when KL divergence falls below this threshold

• min_steps (int, default: 50) – Minimum number of steps before early stopping

• window_size (int, default: 5) – Number of steps to average KL over

__init__(d3pm, kl_threshold=0.01, min_steps=50, window_size=5)

Parameters:

• d3pm (D3PM)

• kl_threshold (float, default: 0.01)

• min_steps (int, default: 50)

• window_size (int, default: 5)

reset()

Reset state for a new sampling run.

Return type:

None

should_stop(step, max_steps, kl_divergence)

Determine if generation should stop based on KL divergence.

Uses tensor operations to avoid GPU->CPU sync in the hot sampling loop.

Parameters:

• step (int) – Current step number (from 0, counting down)

• max_steps (int) – Maximum number of steps

• kl_divergence (Tensor) – KL divergence tensor (scalar) between consecutive predictions

Return type:

bool

Returns:

True if generation should stop, False otherwise

sample(model, shape, temperature=1.0, **model_kwargs)

Generate samples with KL-guided early stopping.

Parameters:

• model (Module) – Trained denoising model

• shape (tuple) – Shape of samples (batch_size, seq_len)

• temperature (float, default: 1.0) – Sampling temperature

• **model_kwargs – Additional model arguments

Returns:

Generated samples metrics: Dict with generation statistics

Return type:

x_0

Continuous Diffusion

class medlatents.diffusion.ContinuousGaussianDiffusion(num_timesteps=1000, schedule_type='cosine', prediction_type='epsilon', device=None, beta_start=0.0001, beta_end=0.02)

Bases: object

Gaussian diffusion process for continuous latent spaces.

Parameters:

• num_timesteps (int, default: 1000)

• schedule_type (str, default: 'cosine')

• prediction_type (str, default: 'epsilon')

• device (device | None, default: None)

• beta_start (float, default: 0.0001)

• beta_end (float, default: 0.02)

__init__(num_timesteps=1000, schedule_type='cosine', prediction_type='epsilon', device=None, beta_start=0.0001, beta_end=0.02)

Parameters:

• num_timesteps (int, default: 1000)

• schedule_type (str, default: 'cosine')

• prediction_type (str, default: 'epsilon')

• device (device | None, default: None)

• beta_start (float, default: 0.0001)

• beta_end (float, default: 0.02)

register_schedule(betas)

Parameters:

betas (Tensor)

Return type:

None

sample_noise_like(x)

Parameters:

x (Tensor)

Return type:

Tensor

q_sample(x_start, t, noise=None)

Sample from q(x_t | x_0) using the reparameterization trick.

The forward diffusion process gradually adds Gaussian noise to the data: q(x_t | x_0) = 𝒩(x_t; √ᾱ_t x_0, (1 - ᾱ_t)I)

where ᾱ_t = ∏ᵢ₌₁ᵗ αᵢ is the cumulative product of noise schedule.

This is efficiently sampled using the reparameterization trick: x_t = √ᾱ_t · x_0 + √(1 - ᾱ_t) · ε, where ε ~ 𝒩(0, I)

Parameters:

• x_start (']) – Clean data x_0

• t (Tensor) – Timestep indices [batch]

• noise ('] | None, default: None) – Optional pre-sampled noise ε

Return type:

']

Returns:

Noisy data x_t at timestep t

compute_loss(model, x_start, y=None, noise=None, model_kwargs=None)

Compute diffusion training loss.

The model learns to predict one of three targets: - “epsilon”: The noise ε that was added (original DDPM) - “x0”: The clean data x_0 directly - “v”: The velocity v_t = √ᾱ_t · ε - √(1 - ᾱ_t) · x_0 (progressive distillation)

Parameters:

• model – Denoising network

• x_start (']) – Clean data x_0

• y (default: None) – Optional conditioning (e.g., class labels)

• noise ('] | None, default: None) – Optional pre-sampled noise

• model_kwargs (dict | None, default: None) – Additional model arguments

Return type:

Tensor

Returns:

MSE loss between model prediction and target

predict_x0(model_output, x_t, t)

Parameters:

• model_output (Tensor)

• x_t (Tensor)

• t (Tensor)

Return type:

Tensor

p_sample(model, x_t, t, y=None, guidance_scale=1.0, null_y=None, model_kwargs=None)

Parameters:

• x_t (Tensor)

• t (Tensor)

• guidance_scale (float, default: 1.0)

• model_kwargs (dict | None, default: None)

Return type:

tuple[Tensor, Tensor, Tensor]

sample(model, shape, y=None, guidance_scale=1.0, null_y=None, num_inference_steps=None, temperature=1.0, model_kwargs=None, initial_noise=None)

Parameters:

• shape (tuple[int, ...])

• guidance_scale (float, default: 1.0)

• num_inference_steps (int | None, default: None)

• temperature (float, default: 1.0)

• model_kwargs (dict | None, default: None)

• initial_noise (Tensor | None, default: None)

Return type:

Tensor

SEDD / MDLM

Score Entropy Discrete Diffusion and Masked Diffusion Language Models.

class medlatents.diffusion.SEDDLoss(vocab_size, noise_schedule='uniform', hybrid_coeff=0.001)

Bases: torch.nn.modules.module.Module

Score Entropy Discrete Diffusion loss.

SEDD models the ratio score: s(x) = log(p(y)/p(x)) for all y ≠ x, instead of modeling the full distribution. This is more efficient because we only need to predict K-1 values instead of K.

The key insight is that for discrete diffusion, we can decompose:

L_SEDD = E_t,x [sum_{y≠x} q(y|x,t) * (s(y,x) - log(q(x|y,t)/q(y|x,t)))^2]

Where s(y,x) = log(p(y)/p(x)) is the ratio score we’re learning.

Reference:

Lou et al., “Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution” (ICML 2024 Best Paper)

Parameters:

• vocab_size (int) – Number of discrete classes

• noise_schedule (str, default: 'uniform') – Type of noise schedule (‘uniform’, ‘absorbing’)

• hybrid_coeff (float, default: 0.001) – Weight for auxiliary cross-entropy loss (default 0.001)

__init__(vocab_size, noise_schedule='uniform', hybrid_coeff=0.001)

Initialize internal Module state, shared by both nn.Module and ScriptModule.

Parameters:

• vocab_size (int)

• noise_schedule (str, default: 'uniform')

• hybrid_coeff (float, default: 0.001)

compute_transition_rates(t, x_t)

Compute transition rates q(y|x,t) for the forward process.

For uniform noise: q(y|x,t) = beta(t)/K for y ≠ x For absorbing: q(y|x,t) = beta(t) * (y == mask)

Parameters:

• t (Tensor) – Time values [batch]

• x_t (Tensor) – Current states [batch, seq_len]

Returns:

[batch, seq_len, vocab_size] transition rates

Return type:

rates

forward(score_logits, x_0, x_t, t)

Compute SEDD loss.

Parameters:

• score_logits (Tensor) – [batch, seq_len, vocab_size] predicted ratio scores where score_logits[…, y] = log(p(y)/p(x)) for y ≠ x

• x_0 (Tensor) – [batch, seq_len] original clean data

• x_t (Tensor) – [batch, seq_len] noisy data at time t

• t (Tensor) – [batch] time values in [0, 1]

Returns:

Scalar SEDD loss

Return type:

loss

T_destination = ~T_destination

add_module(name, module)

Add a child module to the current module.

The module can be accessed as an attribute using the given name.

Parameters:

• name (str) – name of the child module. The child module can be accessed from this module using the given name

• module (Module) – child module to be added to the module.

Return type:

None

apply(fn)

Apply fn recursively to every submodule (as returned by .children()) as well as self.

Typical use includes initializing the parameters of a model (see also torch.nn.init).

Parameters:

fn (Module -> None) – function to be applied to each submodule

Returns:

self

Return type:

Module

Example:

>>> @torch.no_grad()
>>> def init_weights(m):
>>>     print(m)
>>>     if type(m) is nn.Linear:
>>>         m.weight.fill_(1.0)
>>>         print(m.weight)
>>> net = nn.Sequential(nn.Linear(2, 2), nn.Linear(2, 2))
>>> net.apply(init_weights)
Linear(in_features=2, out_features=2, bias=True)
Parameter containing:
tensor([[1., 1.],
        [1., 1.]], requires_grad=True)
Linear(in_features=2, out_features=2, bias=True)
Parameter containing:
tensor([[1., 1.],
        [1., 1.]], requires_grad=True)
Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
)

bfloat16()

Casts all floating point parameters and buffers to bfloat16 datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

buffers(recurse=True)

Return an iterator over module buffers.

Parameters:

recurse (bool) – if True, then yields buffers of this module and all submodules. Otherwise, yields only buffers that are direct members of this module.

Yields:

torch.Tensor – module buffer

Return type:

Iterator[Tensor]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for buf in model.buffers():
>>>     print(type(buf), buf.size())
<class 'torch.Tensor'> (20L,)
<class 'torch.Tensor'> (20L, 1L, 5L, 5L)

call_super_init: bool = False

children()

Return an iterator over immediate children modules.

Yields:

Module – a child module

Return type:

Iterator[Module]

compile(*args, **kwargs)

Compile this Module’s forward using torch.compile().

This Module’s __call__ method is compiled and all arguments are passed as-is to torch.compile().

See torch.compile() for details on the arguments for this function.

Return type:

None

cpu()

Move all model parameters and buffers to the CPU.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

cuda(device=None)

Move all model parameters and buffers to the GPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on GPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

double()

Casts all floating point parameters and buffers to double datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

dump_patches: bool = False

eval()

Set the module in evaluation mode.

This has an effect only on certain modules. See the documentation of particular modules for details of their behaviors in training/evaluation mode, i.e. whether they are affected, e.g. Dropout, BatchNorm, etc.

This is equivalent with self.train(False).

See Locally disabling gradient computation for a comparison between .eval() and several similar mechanisms that may be confused with it.

Returns:

self

Return type:

Module

extra_repr()

Return the extra representation of the module.

To print customized extra information, you should re-implement this method in your own modules. Both single-line and multi-line strings are acceptable.

Return type:

str

float()

Casts all floating point parameters and buffers to float datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

get_buffer(target)

Return the buffer given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed explanation of this method’s functionality as well as how to correctly specify target.

Parameters:

target (str) – The fully-qualified string name of the buffer to look for. (See get_submodule for how to specify a fully-qualified string.)

Returns:

The buffer referenced by target

Return type:

torch.Tensor

Raises:

AttributeError – If the target string references an invalid path or resolves to something that is not a buffer

get_extra_state()

Return any extra state to include in the module’s state_dict.

Implement this and a corresponding set_extra_state() for your module if you need to store extra state. This function is called when building the module’s state_dict().

Note that extra state should be picklable to ensure working serialization of the state_dict. We only provide backwards compatibility guarantees for serializing Tensors; other objects may break backwards compatibility if their serialized pickled form changes.

Returns:

Any extra state to store in the module’s state_dict

Return type:

object

get_parameter(target)

Return the parameter given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed explanation of this method’s functionality as well as how to correctly specify target.

Parameters:

target (str) – The fully-qualified string name of the Parameter to look for. (See get_submodule for how to specify a fully-qualified string.)

Returns:

The Parameter referenced by target

Return type:

torch.nn.Parameter

Raises:

AttributeError – If the target string references an invalid path or resolves to something that is not an nn.Parameter

get_submodule(target)

Return the submodule given by target if it exists, otherwise throw an error.

For example, let’s say you have an nn.Module A that looks like this:

A(
    (net_b): Module(
        (net_c): Module(
            (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))
        )
        (linear): Linear(in_features=100, out_features=200, bias=True)
    )
)

(The diagram shows an nn.Module A. A which has a nested submodule net_b, which itself has two submodules net_c and linear. net_c then has a submodule conv.)

To check whether or not we have the linear submodule, we would call get_submodule("net_b.linear"). To check whether we have the conv submodule, we would call get_submodule("net_b.net_c.conv").

The runtime of get_submodule is bounded by the degree of module nesting in target. A query against named_modules achieves the same result, but it is O(N) in the number of transitive modules. So, for a simple check to see if some submodule exists, get_submodule should always be used.

Parameters:

target (str) – The fully-qualified string name of the submodule to look for. (See above example for how to specify a fully-qualified string.)

Returns:

The submodule referenced by target

Return type:

torch.nn.Module

Raises:

AttributeError – If at any point along the path resulting from the target string the (sub)path resolves to a non-existent attribute name or an object that is not an instance of nn.Module.

half()

Casts all floating point parameters and buffers to half datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

ipu(device=None)

Move all model parameters and buffers to the IPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on IPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

load_state_dict(state_dict, strict=True, assign=False)

Copy parameters and buffers from state_dict into this module and its descendants.

If strict is True, then the keys of state_dict must exactly match the keys returned by this module’s state_dict() function.

Warning

If assign is True the optimizer must be created after the call to load_state_dict unless get_swap_module_params_on_conversion() is True.

Parameters:

• state_dict (dict) – a dict containing parameters and persistent buffers.

• strict (bool, optional) – whether to strictly enforce that the keys in state_dict match the keys returned by this module’s state_dict() function. Default: True

• assign (bool, optional) – When set to False, the properties of the tensors in the current module are preserved whereas setting it to True preserves properties of the Tensors in the state dict. The only exception is the requires_grad field of Parameter for which the value from the module is preserved. Default: False

Returns:

• missing_keys is a list of str containing any keys that are expected

  by this module but missing from the provided state_dict.

• unexpected_keys is a list of str containing the keys that are not

  expected by this module but present in the provided state_dict.

Return type:

NamedTuple with missing_keys and unexpected_keys fields

Note

If a parameter or buffer is registered as None and its corresponding key exists in state_dict, load_state_dict() will raise a RuntimeError.

modules(remove_duplicate=True)

Return an iterator over all modules in the network.

Parameters:

remove_duplicate (bool, default: True) – whether to remove the duplicated module instances in the result or not.

Yields:

Module – a module in the network

Return type:

Iterator[Module]

Note

Duplicate modules are returned only once by default. In the following example, l will be returned only once.

Example:

>>> l = nn.Linear(2, 2)
>>> net = nn.Sequential(l, l)
>>> for idx, m in enumerate(net.modules()):
...     print(idx, '->', m)

0 -> Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
)
1 -> Linear(in_features=2, out_features=2, bias=True)

mtia(device=None)

Move all model parameters and buffers to the MTIA.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on MTIA while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

named_buffers(prefix='', recurse=True, remove_duplicate=True)

Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.

Parameters:

• prefix (str) – prefix to prepend to all buffer names.

• recurse (bool, optional) – if True, then yields buffers of this module and all submodules. Otherwise, yields only buffers that are direct members of this module. Defaults to True.

• remove_duplicate (bool, optional) – whether to remove the duplicated buffers in the result. Defaults to True.

Yields:

(str, torch.Tensor) – Tuple containing the name and buffer

Return type:

Iterator[tuple[str, Tensor]]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, buf in self.named_buffers():
>>>     if name in ['running_var']:
>>>         print(buf.size())

named_children()

Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.

Yields:

(str, Module) – Tuple containing a name and child module

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, module in model.named_children():
>>>     if name in ['conv4', 'conv5']:
>>>         print(module)

Return type:

Iterator[tuple[str, Module]]

named_modules(memo=None, prefix='', remove_duplicate=True)

Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.

Parameters:

• memo (set[Module] | None, default: None) – a memo to store the set of modules already added to the result

• prefix (str, default: '') – a prefix that will be added to the name of the module

• remove_duplicate (bool, default: True) – whether to remove the duplicated module instances in the result or not

Yields:

(str, Module) – Tuple of name and module

Note

Duplicate modules are returned only once. In the following example, l will be returned only once.

Example:

>>> l = nn.Linear(2, 2)
>>> net = nn.Sequential(l, l)
>>> for idx, m in enumerate(net.named_modules()):
...     print(idx, '->', m)

0 -> ('', Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
))
1 -> ('0', Linear(in_features=2, out_features=2, bias=True))

named_parameters(prefix='', recurse=True, remove_duplicate=True)

Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.

Parameters:

• prefix (str) – prefix to prepend to all parameter names.

• recurse (bool) – if True, then yields parameters of this module and all submodules. Otherwise, yields only parameters that are direct members of this module.

• remove_duplicate (bool, optional) – whether to remove the duplicated parameters in the result. Defaults to True.

Yields:

(str, Parameter) – Tuple containing the name and parameter

Return type:

Iterator[tuple[str, Parameter]]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, param in self.named_parameters():
>>>     if name in ['bias']:
>>>         print(param.size())

parameters(recurse=True)

Return an iterator over module parameters.

This is typically passed to an optimizer.

Parameters:

recurse (bool) – if True, then yields parameters of this module and all submodules. Otherwise, yields only parameters that are direct members of this module.

Yields:

Parameter – module parameter

Return type:

Iterator[Parameter]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for param in model.parameters():
>>>     print(type(param), param.size())
<class 'torch.Tensor'> (20L,)
<class 'torch.Tensor'> (20L, 1L, 5L, 5L)

register_backward_hook(hook)

Register a backward hook on the module.

This function is deprecated in favor of register_full_backward_hook() and the behavior of this function will change in future versions.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

Parameters:

hook (Callable[[Module, tuple[Tensor, ...] | Tensor, tuple[Tensor, ...] | Tensor], tuple[Tensor, ...] | Tensor | None])

register_buffer(name, tensor, persistent=True)

Add a buffer to the module.

This is typically used to register a buffer that should not be considered a model parameter. For example, BatchNorm’s running_mean is not a parameter, but is part of the module’s state. Buffers, by default, are persistent and will be saved alongside parameters. This behavior can be changed by setting persistent to False. The only difference between a persistent buffer and a non-persistent buffer is that the latter will not be a part of this module’s state_dict.

Buffers can be accessed as attributes using given names.

Parameters:

• name (str) – name of the buffer. The buffer can be accessed from this module using the given name

• tensor (Tensor or None) – buffer to be registered. If None, then operations that run on buffers, such as cuda, are ignored. If None, the buffer is not included in the module’s state_dict.

• persistent (bool) – whether the buffer is part of this module’s state_dict.

Return type:

None

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> self.register_buffer('running_mean', torch.zeros(num_features))

register_forward_hook(hook, *, prepend=False, with_kwargs=False, always_call=False)

Register a forward hook on the module.

The hook will be called every time after forward() has computed an output.

If with_kwargs is False or not specified, the input contains only the positional arguments given to the module. Keyword arguments won’t be passed to the hooks and only to the forward. The hook can modify the output. It can modify the input inplace but it will not have effect on forward since this is called after forward() is called. The hook should have the following signature:

hook(module, args, output) -> None or modified output

If with_kwargs is True, the forward hook will be passed the kwargs given to the forward function and be expected to return the output possibly modified. The hook should have the following signature:

hook(module, args, kwargs, output) -> None or modified output

Parameters:

• hook (Callable) – The user defined hook to be registered.

• prepend (bool) – If True, the provided hook will be fired before all existing forward hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing forward hooks on this torch.nn.Module. Note that global forward hooks registered with register_module_forward_hook() will fire before all hooks registered by this method. Default: False

• with_kwargs (bool) – If True, the hook will be passed the kwargs given to the forward function. Default: False

• always_call (bool) – If True the hook will be run regardless of whether an exception is raised while calling the Module. Default: False

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_forward_pre_hook(hook, *, prepend=False, with_kwargs=False)

Register a forward pre-hook on the module.

The hook will be called every time before forward() is invoked.

If with_kwargs is false or not specified, the input contains only the positional arguments given to the module. Keyword arguments won’t be passed to the hooks and only to the forward. The hook can modify the input. User can either return a tuple or a single modified value in the hook. We will wrap the value into a tuple if a single value is returned (unless that value is already a tuple). The hook should have the following signature:

hook(module, args) -> None or modified input

If with_kwargs is true, the forward pre-hook will be passed the kwargs given to the forward function. And if the hook modifies the input, both the args and kwargs should be returned. The hook should have the following signature:

hook(module, args, kwargs) -> None or a tuple of modified input and kwargs

Parameters:

• hook (Callable) – The user defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing forward_pre hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing forward_pre hooks on this torch.nn.Module. Note that global forward_pre hooks registered with register_module_forward_pre_hook() will fire before all hooks registered by this method. Default: False

• with_kwargs (bool) – If true, the hook will be passed the kwargs given to the forward function. Default: False

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_full_backward_hook(hook, prepend=False)

Register a backward hook on the module.

The hook will be called every time the gradients with respect to a module are computed, and its firing rules are as follows:

1. Ordinarily, the hook fires when the gradients are computed with respect to the module inputs.

2. If none of the module inputs require gradients, the hook will fire when the gradients are computed with respect to module outputs.

3. If none of the module outputs require gradients, then the hooks will not fire.

The hook should have the following signature:

hook(module, grad_input, grad_output) -> tuple(Tensor) or None

The grad_input and grad_output are tuples that contain the gradients with respect to the inputs and outputs respectively. The hook should not modify its arguments, but it can optionally return a new gradient with respect to the input that will be used in place of grad_input in subsequent computations. grad_input will only correspond to the inputs given as positional arguments and all kwarg arguments are ignored. Entries in grad_input and grad_output will be None for all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will receive a view of each Tensor passed to the Module. Similarly the caller will receive a view of each Tensor returned by the Module’s forward function.

Warning

Modifying inputs or outputs inplace is not allowed when using backward hooks and will raise an error.

Parameters:

• hook (Callable) – The user-defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing backward hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing backward hooks on this torch.nn.Module. Note that global backward hooks registered with register_module_full_backward_hook() will fire before all hooks registered by this method.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_full_backward_pre_hook(hook, prepend=False)

Register a backward pre-hook on the module.

The hook will be called every time the gradients for the module are computed. The hook should have the following signature:

hook(module, grad_output) -> tuple[Tensor, ...], Tensor or None

The grad_output is a tuple. The hook should not modify its arguments, but it can optionally return a new gradient with respect to the output that will be used in place of grad_output in subsequent computations. Entries in grad_output will be None for all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will receive a view of each Tensor passed to the Module. Similarly the caller will receive a view of each Tensor returned by the Module’s forward function.

Warning

Modifying inputs inplace is not allowed when using backward hooks and will raise an error.

Parameters:

• hook (Callable) – The user-defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing backward_pre hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing backward_pre hooks on this torch.nn.Module. Note that global backward_pre hooks registered with register_module_full_backward_pre_hook() will fire before all hooks registered by this method.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_load_state_dict_post_hook(hook)

Register a post-hook to be run after module’s load_state_dict() is called.

It should have the following signature::

hook(module, incompatible_keys) -> None

The module argument is the current module that this hook is registered on, and the incompatible_keys argument is a NamedTuple consisting of attributes missing_keys and unexpected_keys. missing_keys is a list of str containing the missing keys and unexpected_keys is a list of str containing the unexpected keys.

The given incompatible_keys can be modified inplace if needed.

Note that the checks performed when calling load_state_dict() with strict=True are affected by modifications the hook makes to missing_keys or unexpected_keys, as expected. Additions to either set of keys will result in an error being thrown when strict=True, and clearing out both missing and unexpected keys will avoid an error.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_load_state_dict_pre_hook(hook)

Register a pre-hook to be run before module’s load_state_dict() is called.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs) -> None # noqa: B950

Parameters:

hook (Callable) – Callable hook that will be invoked before loading the state dict.

register_module(name, module)

Alias for add_module().

Parameters:

• name (str)

• module (Optional[Module])

Return type:

None

register_parameter(name, param)

Add a parameter to the module.

The parameter can be accessed as an attribute using given name.

Parameters:

• name (str) – name of the parameter. The parameter can be accessed from this module using the given name

• param (Parameter or None) – parameter to be added to the module. If None, then operations that run on parameters, such as cuda, are ignored. If None, the parameter is not included in the module’s state_dict.

Return type:

None

register_state_dict_post_hook(hook)

Register a post-hook for the state_dict() method.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata) -> None

The registered hooks can modify the state_dict inplace.

register_state_dict_pre_hook(hook)

Register a pre-hook for the state_dict() method.

It should have the following signature::

hook(module, prefix, keep_vars) -> None

The registered hooks can be used to perform pre-processing before the state_dict call is made.

requires_grad_(requires_grad=True)

Change if autograd should record operations on parameters in this module.

This method sets the parameters’ requires_grad attributes in-place.

This method is helpful for freezing part of the module for finetuning or training parts of a model individually (e.g., GAN training).

See Locally disabling gradient computation for a comparison between .requires_grad_() and several similar mechanisms that may be confused with it.

Parameters:

requires_grad (bool) – whether autograd should record operations on parameters in this module. Default: True.

Returns:

self

Return type:

Module

set_extra_state(state)

Set extra state contained in the loaded state_dict.

This function is called from load_state_dict() to handle any extra state found within the state_dict. Implement this function and a corresponding get_extra_state() for your module if you need to store extra state within its state_dict.

Parameters:

state (dict) – Extra state from the state_dict

Return type:

None

set_submodule(target, module, strict=False)

Set the submodule given by target if it exists, otherwise throw an error.

Note

If strict is set to False (default), the method will replace an existing submodule or create a new submodule if the parent module exists. If strict is set to True, the method will only attempt to replace an existing submodule and throw an error if the submodule does not exist.

For example, let’s say you have an nn.Module A that looks like this:

A(
    (net_b): Module(
        (net_c): Module(
            (conv): Conv2d(3, 3, 3)
        )
        (linear): Linear(3, 3)
    )
)

(The diagram shows an nn.Module A. A has a nested submodule net_b, which itself has two submodules net_c and linear. net_c then has a submodule conv.)

To override the Conv2d with a new submodule Linear, you could call set_submodule("net_b.net_c.conv", nn.Linear(1, 1)) where strict could be True or False

To add a new submodule Conv2d to the existing net_b module, you would call set_submodule("net_b.conv", nn.Conv2d(1, 1, 1)).

In the above if you set strict=True and call set_submodule("net_b.conv", nn.Conv2d(1, 1, 1), strict=True), an AttributeError will be raised because net_b does not have a submodule named conv.

Parameters:

• target (str) – The fully-qualified string name of the submodule to look for. (See above example for how to specify a fully-qualified string.)

• module (Module) – The module to set the submodule to.

• strict (bool, default: False) – If False, the method will replace an existing submodule or create a new submodule if the parent module exists. If True, the method will only attempt to replace an existing submodule and throw an error if the submodule doesn’t already exist.

Raises:

• ValueError – If the target string is empty or if module is not an instance of nn.Module.

• AttributeError – If at any point along the path resulting from the target string the (sub)path resolves to a non-existent attribute name or an object that is not an instance of nn.Module.

Return type:

None

share_memory()

See torch.Tensor.share_memory_().

Return type:

Self

state_dict(*args, destination=None, prefix='', keep_vars=False)

Return a dictionary containing references to the whole state of the module.

Both parameters and persistent buffers (e.g. running averages) are included. Keys are corresponding parameter and buffer names. Parameters and buffers set to None are not included.

Note

The returned object is a shallow copy. It contains references to the module’s parameters and buffers.

Warning

Currently state_dict() also accepts positional arguments for destination, prefix and keep_vars in order. However, this is being deprecated and keyword arguments will be enforced in future releases.

Warning

Please avoid the use of argument destination as it is not designed for end-users.

Parameters:

• destination (dict, optional) – If provided, the state of module will be updated into the dict and the same object is returned. Otherwise, an OrderedDict will be created and returned. Default: None.

• prefix (str, optional) – a prefix added to parameter and buffer names to compose the keys in state_dict. Default: ''.

• keep_vars (bool, optional) – by default the Tensor s returned in the state dict are detached from autograd. If it’s set to True, detaching will not be performed. Default: False.

Returns:

a dictionary containing a whole state of the module

Return type:

dict

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> module.state_dict().keys()
['bias', 'weight']

to(*args, **kwargs)

Move and/or cast the parameters and buffers.

This can be called as

to(device=None, dtype=None, non_blocking=False)

to(dtype, non_blocking=False)

to(tensor, non_blocking=False)

to(memory_format=torch.channels_last)

Its signature is similar to torch.Tensor.to(), but only accepts floating point or complex dtypes. In addition, this method will only cast the floating point or complex parameters and buffers to dtype (if given). The integral parameters and buffers will be moved device, if that is given, but with dtypes unchanged. When non_blocking is set, it tries to convert/move asynchronously with respect to the host if possible, e.g., moving CPU Tensors with pinned memory to CUDA devices.

See below for examples.

Note

This method modifies the module in-place.

Parameters:

• device (torch.device) – the desired device of the parameters and buffers in this module

• dtype (torch.dtype) – the desired floating point or complex dtype of the parameters and buffers in this module

• tensor (torch.Tensor) – Tensor whose dtype and device are the desired dtype and device for all parameters and buffers in this module

• memory_format (torch.memory_format) – the desired memory format for 4D parameters and buffers in this module (keyword only argument)

Returns:

self

Return type:

Module

Examples:

>>> # xdoctest: +IGNORE_WANT("non-deterministic")
>>> linear = nn.Linear(2, 2)
>>> linear.weight
Parameter containing:
tensor([[ 0.1913, -0.3420],
        [-0.5113, -0.2325]])
>>> linear.to(torch.double)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1913, -0.3420],
        [-0.5113, -0.2325]], dtype=torch.float64)
>>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA1)
>>> gpu1 = torch.device("cuda:1")
>>> linear.to(gpu1, dtype=torch.half, non_blocking=True)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1914, -0.3420],
        [-0.5112, -0.2324]], dtype=torch.float16, device='cuda:1')
>>> cpu = torch.device("cpu")
>>> linear.to(cpu)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1914, -0.3420],
        [-0.5112, -0.2324]], dtype=torch.float16)

>>> linear = nn.Linear(2, 2, bias=None).to(torch.cdouble)
>>> linear.weight
Parameter containing:
tensor([[ 0.3741+0.j,  0.2382+0.j],
        [ 0.5593+0.j, -0.4443+0.j]], dtype=torch.complex128)
>>> linear(torch.ones(3, 2, dtype=torch.cdouble))
tensor([[0.6122+0.j, 0.1150+0.j],
        [0.6122+0.j, 0.1150+0.j],
        [0.6122+0.j, 0.1150+0.j]], dtype=torch.complex128)

to_empty(*, device, recurse=True)

Move the parameters and buffers to the specified device without copying storage.

Parameters:

• device (torch.device) – The desired device of the parameters and buffers in this module.

• recurse (bool) – Whether parameters and buffers of submodules should be recursively moved to the specified device.

Returns:

self

Return type:

Module

train(mode=True)

Set the module in training mode.

This has an effect only on certain modules. See the documentation of particular modules for details of their behaviors in training/evaluation mode, i.e., whether they are affected, e.g. Dropout, BatchNorm, etc.

Parameters:

mode (bool) – whether to set training mode (True) or evaluation mode (False). Default: True.

Returns:

self

Return type:

Module

type(dst_type)

Casts all parameters and buffers to dst_type.

Note

This method modifies the module in-place.

Parameters:

dst_type (type or string) – the desired type

Returns:

self

Return type:

Module

xpu(device=None)

Move all model parameters and buffers to the XPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing optimizer if the module will live on XPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

zero_grad(set_to_none=True)

Reset gradients of all model parameters.

See similar function under torch.optim.Optimizer for more context.

Parameters:

set_to_none (bool) – instead of setting to zero, set the grads to None. See torch.optim.Optimizer.zero_grad() for details.

Return type:

None

training: bool

class medlatents.diffusion.MDLMLoss(vocab_size, mask_token=None, time_reweighting='uniform')

Bases: torch.nn.modules.module.Module

Masked Diffusion Language Model (MDLM) Rao-Blackwellized objective.

MDLM provides a variance-reduced objective for masked diffusion by analytically marginalizing out the discrete sampling in the ELBO.

Key insight: Instead of sampling x_t and computing the loss, we can directly compute the expected loss over all possible x_t values, weighted by their probabilities under the forward process.

This Rao-Blackwellization eliminates variance from discrete sampling, leading to faster and more stable training.

Reference:

Sahoo et al., “Simple and Effective Masked Diffusion Language Models” (arXiv:2406.07524)

Parameters:

• vocab_size (int) – Number of discrete classes

• mask_token (int | None, default: None) – Index of mask token (default: vocab_size - 1)

• time_reweighting (str, default: 'uniform') – Type of time reweighting (‘uniform’, ‘importance’, ‘snr’)

__init__(vocab_size, mask_token=None, time_reweighting='uniform')

Initialize internal Module state, shared by both nn.Module and ScriptModule.

Parameters:

• vocab_size (int)

• mask_token (int | None, default: None)

• time_reweighting (str, default: 'uniform')

get_time_weights(t)

Get importance weights for different time values.

Parameters:

t (Tensor) – Time values [batch] in [0, 1]

Returns:

[batch] importance weights

Return type:

weights

forward(logits, x_0, t)

Compute MDLM Rao-Blackwellized loss.

Parameters:

• logits (Tensor) – [batch, seq_len, vocab_size] predicted logits

• x_0 (Tensor) – [batch, seq_len] clean data

• t (Tensor) – [batch] time values in [0, 1]

Returns:

Scalar MDLM loss

Return type:

loss

T_destination = ~T_destination

add_module(name, module)

Add a child module to the current module.

The module can be accessed as an attribute using the given name.

Parameters:

• name (str) – name of the child module. The child module can be accessed from this module using the given name

• module (Module) – child module to be added to the module.

Return type:

None

apply(fn)

Apply fn recursively to every submodule (as returned by .children()) as well as self.

Typical use includes initializing the parameters of a model (see also torch.nn.init).

Parameters:

fn (Module -> None) – function to be applied to each submodule

Returns:

self

Return type:

Module

Example:

>>> @torch.no_grad()
>>> def init_weights(m):
>>>     print(m)
>>>     if type(m) is nn.Linear:
>>>         m.weight.fill_(1.0)
>>>         print(m.weight)
>>> net = nn.Sequential(nn.Linear(2, 2), nn.Linear(2, 2))
>>> net.apply(init_weights)
Linear(in_features=2, out_features=2, bias=True)
Parameter containing:
tensor([[1., 1.],
        [1., 1.]], requires_grad=True)
Linear(in_features=2, out_features=2, bias=True)
Parameter containing:
tensor([[1., 1.],
        [1., 1.]], requires_grad=True)
Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
)

bfloat16()

Casts all floating point parameters and buffers to bfloat16 datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

buffers(recurse=True)

Return an iterator over module buffers.

Parameters:

recurse (bool) – if True, then yields buffers of this module and all submodules. Otherwise, yields only buffers that are direct members of this module.

Yields:

torch.Tensor – module buffer

Return type:

Iterator[Tensor]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for buf in model.buffers():
>>>     print(type(buf), buf.size())
<class 'torch.Tensor'> (20L,)
<class 'torch.Tensor'> (20L, 1L, 5L, 5L)

call_super_init: bool = False

children()

Return an iterator over immediate children modules.

Yields:

Module – a child module

Return type:

Iterator[Module]

compile(*args, **kwargs)

Compile this Module’s forward using torch.compile().

This Module’s __call__ method is compiled and all arguments are passed as-is to torch.compile().

See torch.compile() for details on the arguments for this function.

Return type:

None

cpu()

Move all model parameters and buffers to the CPU.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

cuda(device=None)

Move all model parameters and buffers to the GPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on GPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

double()

Casts all floating point parameters and buffers to double datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

dump_patches: bool = False

eval()

Set the module in evaluation mode.

This has an effect only on certain modules. See the documentation of particular modules for details of their behaviors in training/evaluation mode, i.e. whether they are affected, e.g. Dropout, BatchNorm, etc.

This is equivalent with self.train(False).

See Locally disabling gradient computation for a comparison between .eval() and several similar mechanisms that may be confused with it.

Returns:

self

Return type:

Module

extra_repr()

Return the extra representation of the module.

To print customized extra information, you should re-implement this method in your own modules. Both single-line and multi-line strings are acceptable.

Return type:

str

float()

Casts all floating point parameters and buffers to float datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

get_buffer(target)

Return the buffer given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed explanation of this method’s functionality as well as how to correctly specify target.

Parameters:

target (str) – The fully-qualified string name of the buffer to look for. (See get_submodule for how to specify a fully-qualified string.)

Returns:

The buffer referenced by target

Return type:

torch.Tensor

Raises:

AttributeError – If the target string references an invalid path or resolves to something that is not a buffer

get_extra_state()

Return any extra state to include in the module’s state_dict.

Implement this and a corresponding set_extra_state() for your module if you need to store extra state. This function is called when building the module’s state_dict().

Note that extra state should be picklable to ensure working serialization of the state_dict. We only provide backwards compatibility guarantees for serializing Tensors; other objects may break backwards compatibility if their serialized pickled form changes.

Returns:

Any extra state to store in the module’s state_dict

Return type:

object

get_parameter(target)

Return the parameter given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed explanation of this method’s functionality as well as how to correctly specify target.

Parameters:

target (str) – The fully-qualified string name of the Parameter to look for. (See get_submodule for how to specify a fully-qualified string.)

Returns:

The Parameter referenced by target

Return type:

torch.nn.Parameter

Raises:

AttributeError – If the target string references an invalid path or resolves to something that is not an nn.Parameter

get_submodule(target)

Return the submodule given by target if it exists, otherwise throw an error.

For example, let’s say you have an nn.Module A that looks like this:

A(
    (net_b): Module(
        (net_c): Module(
            (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))
        )
        (linear): Linear(in_features=100, out_features=200, bias=True)
    )
)

(The diagram shows an nn.Module A. A which has a nested submodule net_b, which itself has two submodules net_c and linear. net_c then has a submodule conv.)

To check whether or not we have the linear submodule, we would call get_submodule("net_b.linear"). To check whether we have the conv submodule, we would call get_submodule("net_b.net_c.conv").

The runtime of get_submodule is bounded by the degree of module nesting in target. A query against named_modules achieves the same result, but it is O(N) in the number of transitive modules. So, for a simple check to see if some submodule exists, get_submodule should always be used.

Parameters:

target (str) – The fully-qualified string name of the submodule to look for. (See above example for how to specify a fully-qualified string.)

Returns:

The submodule referenced by target

Return type:

torch.nn.Module

Raises:

AttributeError – If at any point along the path resulting from the target string the (sub)path resolves to a non-existent attribute name or an object that is not an instance of nn.Module.

half()

Casts all floating point parameters and buffers to half datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

ipu(device=None)

Move all model parameters and buffers to the IPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on IPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

load_state_dict(state_dict, strict=True, assign=False)

Copy parameters and buffers from state_dict into this module and its descendants.

If strict is True, then the keys of state_dict must exactly match the keys returned by this module’s state_dict() function.

Warning

If assign is True the optimizer must be created after the call to load_state_dict unless get_swap_module_params_on_conversion() is True.

Parameters:

• state_dict (dict) – a dict containing parameters and persistent buffers.

• strict (bool, optional) – whether to strictly enforce that the keys in state_dict match the keys returned by this module’s state_dict() function. Default: True

• assign (bool, optional) – When set to False, the properties of the tensors in the current module are preserved whereas setting it to True preserves properties of the Tensors in the state dict. The only exception is the requires_grad field of Parameter for which the value from the module is preserved. Default: False

Returns:

• missing_keys is a list of str containing any keys that are expected

  by this module but missing from the provided state_dict.

• unexpected_keys is a list of str containing the keys that are not

  expected by this module but present in the provided state_dict.

Return type:

NamedTuple with missing_keys and unexpected_keys fields

Note

If a parameter or buffer is registered as None and its corresponding key exists in state_dict, load_state_dict() will raise a RuntimeError.

modules(remove_duplicate=True)

Return an iterator over all modules in the network.

Parameters:

remove_duplicate (bool, default: True) – whether to remove the duplicated module instances in the result or not.

Yields:

Module – a module in the network

Return type:

Iterator[Module]

Note

Duplicate modules are returned only once by default. In the following example, l will be returned only once.

Example:

>>> l = nn.Linear(2, 2)
>>> net = nn.Sequential(l, l)
>>> for idx, m in enumerate(net.modules()):
...     print(idx, '->', m)

0 -> Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
)
1 -> Linear(in_features=2, out_features=2, bias=True)

mtia(device=None)

Move all model parameters and buffers to the MTIA.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on MTIA while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

named_buffers(prefix='', recurse=True, remove_duplicate=True)

Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.

Parameters:

• prefix (str) – prefix to prepend to all buffer names.

• recurse (bool, optional) – if True, then yields buffers of this module and all submodules. Otherwise, yields only buffers that are direct members of this module. Defaults to True.

• remove_duplicate (bool, optional) – whether to remove the duplicated buffers in the result. Defaults to True.

Yields:

(str, torch.Tensor) – Tuple containing the name and buffer

Return type:

Iterator[tuple[str, Tensor]]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, buf in self.named_buffers():
>>>     if name in ['running_var']:
>>>         print(buf.size())

named_children()

Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.

Yields:

(str, Module) – Tuple containing a name and child module

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, module in model.named_children():
>>>     if name in ['conv4', 'conv5']:
>>>         print(module)

Return type:

Iterator[tuple[str, Module]]

named_modules(memo=None, prefix='', remove_duplicate=True)

Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.

Parameters:

• memo (set[Module] | None, default: None) – a memo to store the set of modules already added to the result

• prefix (str, default: '') – a prefix that will be added to the name of the module

• remove_duplicate (bool, default: True) – whether to remove the duplicated module instances in the result or not

Yields:

(str, Module) – Tuple of name and module

Note

Duplicate modules are returned only once. In the following example, l will be returned only once.

Example:

>>> l = nn.Linear(2, 2)
>>> net = nn.Sequential(l, l)
>>> for idx, m in enumerate(net.named_modules()):
...     print(idx, '->', m)

0 -> ('', Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
))
1 -> ('0', Linear(in_features=2, out_features=2, bias=True))

named_parameters(prefix='', recurse=True, remove_duplicate=True)

Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.

Parameters:

• prefix (str) – prefix to prepend to all parameter names.

• recurse (bool) – if True, then yields parameters of this module and all submodules. Otherwise, yields only parameters that are direct members of this module.

• remove_duplicate (bool, optional) – whether to remove the duplicated parameters in the result. Defaults to True.

Yields:

(str, Parameter) – Tuple containing the name and parameter

Return type:

Iterator[tuple[str, Parameter]]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, param in self.named_parameters():
>>>     if name in ['bias']:
>>>         print(param.size())

parameters(recurse=True)

Return an iterator over module parameters.

This is typically passed to an optimizer.

Parameters:

recurse (bool) – if True, then yields parameters of this module and all submodules. Otherwise, yields only parameters that are direct members of this module.

Yields:

Parameter – module parameter

Return type:

Iterator[Parameter]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for param in model.parameters():
>>>     print(type(param), param.size())
<class 'torch.Tensor'> (20L,)
<class 'torch.Tensor'> (20L, 1L, 5L, 5L)

register_backward_hook(hook)

Register a backward hook on the module.

This function is deprecated in favor of register_full_backward_hook() and the behavior of this function will change in future versions.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

Parameters:

hook (Callable[[Module, tuple[Tensor, ...] | Tensor, tuple[Tensor, ...] | Tensor], tuple[Tensor, ...] | Tensor | None])

register_buffer(name, tensor, persistent=True)

Add a buffer to the module.

This is typically used to register a buffer that should not be considered a model parameter. For example, BatchNorm’s running_mean is not a parameter, but is part of the module’s state. Buffers, by default, are persistent and will be saved alongside parameters. This behavior can be changed by setting persistent to False. The only difference between a persistent buffer and a non-persistent buffer is that the latter will not be a part of this module’s state_dict.

Buffers can be accessed as attributes using given names.

Parameters:

• name (str) – name of the buffer. The buffer can be accessed from this module using the given name

• tensor (Tensor or None) – buffer to be registered. If None, then operations that run on buffers, such as cuda, are ignored. If None, the buffer is not included in the module’s state_dict.

• persistent (bool) – whether the buffer is part of this module’s state_dict.

Return type:

None

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> self.register_buffer('running_mean', torch.zeros(num_features))

register_forward_hook(hook, *, prepend=False, with_kwargs=False, always_call=False)

Register a forward hook on the module.

The hook will be called every time after forward() has computed an output.

If with_kwargs is False or not specified, the input contains only the positional arguments given to the module. Keyword arguments won’t be passed to the hooks and only to the forward. The hook can modify the output. It can modify the input inplace but it will not have effect on forward since this is called after forward() is called. The hook should have the following signature:

hook(module, args, output) -> None or modified output

If with_kwargs is True, the forward hook will be passed the kwargs given to the forward function and be expected to return the output possibly modified. The hook should have the following signature:

hook(module, args, kwargs, output) -> None or modified output

Parameters:

• hook (Callable) – The user defined hook to be registered.

• prepend (bool) – If True, the provided hook will be fired before all existing forward hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing forward hooks on this torch.nn.Module. Note that global forward hooks registered with register_module_forward_hook() will fire before all hooks registered by this method. Default: False

• with_kwargs (bool) – If True, the hook will be passed the kwargs given to the forward function. Default: False

• always_call (bool) – If True the hook will be run regardless of whether an exception is raised while calling the Module. Default: False

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_forward_pre_hook(hook, *, prepend=False, with_kwargs=False)

Register a forward pre-hook on the module.

The hook will be called every time before forward() is invoked.

If with_kwargs is false or not specified, the input contains only the positional arguments given to the module. Keyword arguments won’t be passed to the hooks and only to the forward. The hook can modify the input. User can either return a tuple or a single modified value in the hook. We will wrap the value into a tuple if a single value is returned (unless that value is already a tuple). The hook should have the following signature:

hook(module, args) -> None or modified input

If with_kwargs is true, the forward pre-hook will be passed the kwargs given to the forward function. And if the hook modifies the input, both the args and kwargs should be returned. The hook should have the following signature:

hook(module, args, kwargs) -> None or a tuple of modified input and kwargs

Parameters:

• hook (Callable) – The user defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing forward_pre hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing forward_pre hooks on this torch.nn.Module. Note that global forward_pre hooks registered with register_module_forward_pre_hook() will fire before all hooks registered by this method. Default: False

• with_kwargs (bool) – If true, the hook will be passed the kwargs given to the forward function. Default: False

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_full_backward_hook(hook, prepend=False)

Register a backward hook on the module.

The hook will be called every time the gradients with respect to a module are computed, and its firing rules are as follows:

1. Ordinarily, the hook fires when the gradients are computed with respect to the module inputs.

2. If none of the module inputs require gradients, the hook will fire when the gradients are computed with respect to module outputs.

3. If none of the module outputs require gradients, then the hooks will not fire.

The hook should have the following signature:

hook(module, grad_input, grad_output) -> tuple(Tensor) or None

The grad_input and grad_output are tuples that contain the gradients with respect to the inputs and outputs respectively. The hook should not modify its arguments, but it can optionally return a new gradient with respect to the input that will be used in place of grad_input in subsequent computations. grad_input will only correspond to the inputs given as positional arguments and all kwarg arguments are ignored. Entries in grad_input and grad_output will be None for all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will receive a view of each Tensor passed to the Module. Similarly the caller will receive a view of each Tensor returned by the Module’s forward function.

Warning

Modifying inputs or outputs inplace is not allowed when using backward hooks and will raise an error.

Parameters:

• hook (Callable) – The user-defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing backward hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing backward hooks on this torch.nn.Module. Note that global backward hooks registered with register_module_full_backward_hook() will fire before all hooks registered by this method.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_full_backward_pre_hook(hook, prepend=False)

Register a backward pre-hook on the module.

The hook will be called every time the gradients for the module are computed. The hook should have the following signature:

hook(module, grad_output) -> tuple[Tensor, ...], Tensor or None

The grad_output is a tuple. The hook should not modify its arguments, but it can optionally return a new gradient with respect to the output that will be used in place of grad_output in subsequent computations. Entries in grad_output will be None for all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will receive a view of each Tensor passed to the Module. Similarly the caller will receive a view of each Tensor returned by the Module’s forward function.

Warning

Modifying inputs inplace is not allowed when using backward hooks and will raise an error.

Parameters:

• hook (Callable) – The user-defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing backward_pre hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing backward_pre hooks on this torch.nn.Module. Note that global backward_pre hooks registered with register_module_full_backward_pre_hook() will fire before all hooks registered by this method.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_load_state_dict_post_hook(hook)

Register a post-hook to be run after module’s load_state_dict() is called.

It should have the following signature::

hook(module, incompatible_keys) -> None

The module argument is the current module that this hook is registered on, and the incompatible_keys argument is a NamedTuple consisting of attributes missing_keys and unexpected_keys. missing_keys is a list of str containing the missing keys and unexpected_keys is a list of str containing the unexpected keys.

The given incompatible_keys can be modified inplace if needed.

Note that the checks performed when calling load_state_dict() with strict=True are affected by modifications the hook makes to missing_keys or unexpected_keys, as expected. Additions to either set of keys will result in an error being thrown when strict=True, and clearing out both missing and unexpected keys will avoid an error.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_load_state_dict_pre_hook(hook)

Register a pre-hook to be run before module’s load_state_dict() is called.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs) -> None # noqa: B950

Parameters:

hook (Callable) – Callable hook that will be invoked before loading the state dict.

register_module(name, module)

Alias for add_module().

Parameters:

• name (str)

• module (Optional[Module])

Return type:

None

register_parameter(name, param)

Add a parameter to the module.

The parameter can be accessed as an attribute using given name.

Parameters:

• name (str) – name of the parameter. The parameter can be accessed from this module using the given name

• param (Parameter or None) – parameter to be added to the module. If None, then operations that run on parameters, such as cuda, are ignored. If None, the parameter is not included in the module’s state_dict.

Return type:

None

register_state_dict_post_hook(hook)

Register a post-hook for the state_dict() method.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata) -> None

The registered hooks can modify the state_dict inplace.

register_state_dict_pre_hook(hook)

Register a pre-hook for the state_dict() method.

It should have the following signature::

hook(module, prefix, keep_vars) -> None

The registered hooks can be used to perform pre-processing before the state_dict call is made.

requires_grad_(requires_grad=True)

Change if autograd should record operations on parameters in this module.

This method sets the parameters’ requires_grad attributes in-place.

This method is helpful for freezing part of the module for finetuning or training parts of a model individually (e.g., GAN training).

See Locally disabling gradient computation for a comparison between .requires_grad_() and several similar mechanisms that may be confused with it.

Parameters:

requires_grad (bool) – whether autograd should record operations on parameters in this module. Default: True.

Returns:

self

Return type:

Module

set_extra_state(state)

Set extra state contained in the loaded state_dict.

This function is called from load_state_dict() to handle any extra state found within the state_dict. Implement this function and a corresponding get_extra_state() for your module if you need to store extra state within its state_dict.

Parameters:

state (dict) – Extra state from the state_dict

Return type:

None

set_submodule(target, module, strict=False)

Set the submodule given by target if it exists, otherwise throw an error.

Note

If strict is set to False (default), the method will replace an existing submodule or create a new submodule if the parent module exists. If strict is set to True, the method will only attempt to replace an existing submodule and throw an error if the submodule does not exist.

For example, let’s say you have an nn.Module A that looks like this:

A(
    (net_b): Module(
        (net_c): Module(
            (conv): Conv2d(3, 3, 3)
        )
        (linear): Linear(3, 3)
    )
)

(The diagram shows an nn.Module A. A has a nested submodule net_b, which itself has two submodules net_c and linear. net_c then has a submodule conv.)

To override the Conv2d with a new submodule Linear, you could call set_submodule("net_b.net_c.conv", nn.Linear(1, 1)) where strict could be True or False

To add a new submodule Conv2d to the existing net_b module, you would call set_submodule("net_b.conv", nn.Conv2d(1, 1, 1)).

In the above if you set strict=True and call set_submodule("net_b.conv", nn.Conv2d(1, 1, 1), strict=True), an AttributeError will be raised because net_b does not have a submodule named conv.

Parameters:

• target (str) – The fully-qualified string name of the submodule to look for. (See above example for how to specify a fully-qualified string.)

• module (Module) – The module to set the submodule to.

• strict (bool, default: False) – If False, the method will replace an existing submodule or create a new submodule if the parent module exists. If True, the method will only attempt to replace an existing submodule and throw an error if the submodule doesn’t already exist.

Raises:

• ValueError – If the target string is empty or if module is not an instance of nn.Module.

• AttributeError – If at any point along the path resulting from the target string the (sub)path resolves to a non-existent attribute name or an object that is not an instance of nn.Module.

Return type:

None

share_memory()

See torch.Tensor.share_memory_().

Return type:

Self

state_dict(*args, destination=None, prefix='', keep_vars=False)

Return a dictionary containing references to the whole state of the module.

Both parameters and persistent buffers (e.g. running averages) are included. Keys are corresponding parameter and buffer names. Parameters and buffers set to None are not included.

Note

The returned object is a shallow copy. It contains references to the module’s parameters and buffers.

Warning

Currently state_dict() also accepts positional arguments for destination, prefix and keep_vars in order. However, this is being deprecated and keyword arguments will be enforced in future releases.

Warning

Please avoid the use of argument destination as it is not designed for end-users.

Parameters:

• destination (dict, optional) – If provided, the state of module will be updated into the dict and the same object is returned. Otherwise, an OrderedDict will be created and returned. Default: None.

• prefix (str, optional) – a prefix added to parameter and buffer names to compose the keys in state_dict. Default: ''.

• keep_vars (bool, optional) – by default the Tensor s returned in the state dict are detached from autograd. If it’s set to True, detaching will not be performed. Default: False.

Returns:

a dictionary containing a whole state of the module

Return type:

dict

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> module.state_dict().keys()
['bias', 'weight']

to(*args, **kwargs)

Move and/or cast the parameters and buffers.

This can be called as

to(device=None, dtype=None, non_blocking=False)

to(dtype, non_blocking=False)

to(tensor, non_blocking=False)

to(memory_format=torch.channels_last)

Its signature is similar to torch.Tensor.to(), but only accepts floating point or complex dtypes. In addition, this method will only cast the floating point or complex parameters and buffers to dtype (if given). The integral parameters and buffers will be moved device, if that is given, but with dtypes unchanged. When non_blocking is set, it tries to convert/move asynchronously with respect to the host if possible, e.g., moving CPU Tensors with pinned memory to CUDA devices.

See below for examples.

Note

This method modifies the module in-place.

Parameters:

• device (torch.device) – the desired device of the parameters and buffers in this module

• dtype (torch.dtype) – the desired floating point or complex dtype of the parameters and buffers in this module

• tensor (torch.Tensor) – Tensor whose dtype and device are the desired dtype and device for all parameters and buffers in this module

• memory_format (torch.memory_format) – the desired memory format for 4D parameters and buffers in this module (keyword only argument)

Returns:

self

Return type:

Module

Examples:

>>> # xdoctest: +IGNORE_WANT("non-deterministic")
>>> linear = nn.Linear(2, 2)
>>> linear.weight
Parameter containing:
tensor([[ 0.1913, -0.3420],
        [-0.5113, -0.2325]])
>>> linear.to(torch.double)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1913, -0.3420],
        [-0.5113, -0.2325]], dtype=torch.float64)
>>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA1)
>>> gpu1 = torch.device("cuda:1")
>>> linear.to(gpu1, dtype=torch.half, non_blocking=True)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1914, -0.3420],
        [-0.5112, -0.2324]], dtype=torch.float16, device='cuda:1')
>>> cpu = torch.device("cpu")
>>> linear.to(cpu)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1914, -0.3420],
        [-0.5112, -0.2324]], dtype=torch.float16)

>>> linear = nn.Linear(2, 2, bias=None).to(torch.cdouble)
>>> linear.weight
Parameter containing:
tensor([[ 0.3741+0.j,  0.2382+0.j],
        [ 0.5593+0.j, -0.4443+0.j]], dtype=torch.complex128)
>>> linear(torch.ones(3, 2, dtype=torch.cdouble))
tensor([[0.6122+0.j, 0.1150+0.j],
        [0.6122+0.j, 0.1150+0.j],
        [0.6122+0.j, 0.1150+0.j]], dtype=torch.complex128)

to_empty(*, device, recurse=True)

Move the parameters and buffers to the specified device without copying storage.

Parameters:

• device (torch.device) – The desired device of the parameters and buffers in this module.

• recurse (bool) – Whether parameters and buffers of submodules should be recursively moved to the specified device.

Returns:

self

Return type:

Module

train(mode=True)

Set the module in training mode.

This has an effect only on certain modules. See the documentation of particular modules for details of their behaviors in training/evaluation mode, i.e., whether they are affected, e.g. Dropout, BatchNorm, etc.

Parameters:

mode (bool) – whether to set training mode (True) or evaluation mode (False). Default: True.

Returns:

self

Return type:

Module

type(dst_type)

Casts all parameters and buffers to dst_type.

Note

This method modifies the module in-place.

Parameters:

dst_type (type or string) – the desired type

Returns:

self

Return type:

Module

xpu(device=None)

Move all model parameters and buffers to the XPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing optimizer if the module will live on XPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

zero_grad(set_to_none=True)

Reset gradients of all model parameters.

See similar function under torch.optim.Optimizer for more context.

Parameters:

set_to_none (bool) – instead of setting to zero, set the grads to None. See torch.optim.Optimizer.zero_grad() for details.

Return type:

None

training: bool

class medlatents.diffusion.ContinuousTimeMDLM(vocab_size, mask_token=None, schedule_type='linear')

Bases: torch.nn.modules.module.Module

Continuous-time MDLM with proper ELBO derivation.

Uses continuous-time formulation for tighter bounds and proper handling of the diffusion process.

Reference: Sahoo et al., “Simple and Effective Masked Diffusion Language Models”

Parameters:

• vocab_size (int)

• mask_token (int | None, default: None)

• schedule_type (str, default: 'linear')

__init__(vocab_size, mask_token=None, schedule_type='linear')

Initialize internal Module state, shared by both nn.Module and ScriptModule.

Parameters:

• vocab_size (int)

• mask_token (int | None, default: None)

• schedule_type (str, default: 'linear')

alpha_schedule(t)

Compute alpha(t) = P(not masked at time t).

Parameters:

t (Tensor) – Time values [batch] in [0, 1]

Returns:

[batch] probability of being unmasked

Return type:

alpha

forward(logits, x_0, t)

Compute continuous-time MDLM ELBO.

Parameters:

• logits (Tensor) – [batch, seq_len, vocab_size] predicted logits

• x_0 (Tensor) – [batch, seq_len] clean data

• t (Tensor) – [batch] time values in [0, 1]

Returns:

Scalar ELBO loss

Return type:

loss

T_destination = ~T_destination

add_module(name, module)

Add a child module to the current module.

The module can be accessed as an attribute using the given name.

Parameters:

• name (str) – name of the child module. The child module can be accessed from this module using the given name

• module (Module) – child module to be added to the module.

Return type:

None

apply(fn)

Apply fn recursively to every submodule (as returned by .children()) as well as self.

Typical use includes initializing the parameters of a model (see also torch.nn.init).

Parameters:

fn (Module -> None) – function to be applied to each submodule

Returns:

self

Return type:

Module

Example:

>>> @torch.no_grad()
>>> def init_weights(m):
>>>     print(m)
>>>     if type(m) is nn.Linear:
>>>         m.weight.fill_(1.0)
>>>         print(m.weight)
>>> net = nn.Sequential(nn.Linear(2, 2), nn.Linear(2, 2))
>>> net.apply(init_weights)
Linear(in_features=2, out_features=2, bias=True)
Parameter containing:
tensor([[1., 1.],
        [1., 1.]], requires_grad=True)
Linear(in_features=2, out_features=2, bias=True)
Parameter containing:
tensor([[1., 1.],
        [1., 1.]], requires_grad=True)
Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
)

bfloat16()

Casts all floating point parameters and buffers to bfloat16 datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

buffers(recurse=True)

Return an iterator over module buffers.

Parameters:

recurse (bool) – if True, then yields buffers of this module and all submodules. Otherwise, yields only buffers that are direct members of this module.

Yields:

torch.Tensor – module buffer

Return type:

Iterator[Tensor]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for buf in model.buffers():
>>>     print(type(buf), buf.size())
<class 'torch.Tensor'> (20L,)
<class 'torch.Tensor'> (20L, 1L, 5L, 5L)

call_super_init: bool = False

children()

Return an iterator over immediate children modules.

Yields:

Module – a child module

Return type:

Iterator[Module]

compile(*args, **kwargs)

Compile this Module’s forward using torch.compile().

This Module’s __call__ method is compiled and all arguments are passed as-is to torch.compile().

See torch.compile() for details on the arguments for this function.

Return type:

None

cpu()

Move all model parameters and buffers to the CPU.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

cuda(device=None)

Move all model parameters and buffers to the GPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on GPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

double()

Casts all floating point parameters and buffers to double datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

dump_patches: bool = False

eval()

Set the module in evaluation mode.

This has an effect only on certain modules. See the documentation of particular modules for details of their behaviors in training/evaluation mode, i.e. whether they are affected, e.g. Dropout, BatchNorm, etc.

This is equivalent with self.train(False).

See Locally disabling gradient computation for a comparison between .eval() and several similar mechanisms that may be confused with it.

Returns:

self

Return type:

Module

extra_repr()

Return the extra representation of the module.

To print customized extra information, you should re-implement this method in your own modules. Both single-line and multi-line strings are acceptable.

Return type:

str

float()

Casts all floating point parameters and buffers to float datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

get_buffer(target)

Return the buffer given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed explanation of this method’s functionality as well as how to correctly specify target.

Parameters:

target (str) – The fully-qualified string name of the buffer to look for. (See get_submodule for how to specify a fully-qualified string.)

Returns:

The buffer referenced by target

Return type:

torch.Tensor

Raises:

AttributeError – If the target string references an invalid path or resolves to something that is not a buffer

get_extra_state()

Return any extra state to include in the module’s state_dict.

Implement this and a corresponding set_extra_state() for your module if you need to store extra state. This function is called when building the module’s state_dict().

Note that extra state should be picklable to ensure working serialization of the state_dict. We only provide backwards compatibility guarantees for serializing Tensors; other objects may break backwards compatibility if their serialized pickled form changes.

Returns:

Any extra state to store in the module’s state_dict

Return type:

object

get_parameter(target)

Return the parameter given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed explanation of this method’s functionality as well as how to correctly specify target.

Parameters:

target (str) – The fully-qualified string name of the Parameter to look for. (See get_submodule for how to specify a fully-qualified string.)

Returns:

The Parameter referenced by target

Return type:

torch.nn.Parameter

Raises:

AttributeError – If the target string references an invalid path or resolves to something that is not an nn.Parameter

get_submodule(target)

Return the submodule given by target if it exists, otherwise throw an error.

For example, let’s say you have an nn.Module A that looks like this:

A(
    (net_b): Module(
        (net_c): Module(
            (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))
        )
        (linear): Linear(in_features=100, out_features=200, bias=True)
    )
)

(The diagram shows an nn.Module A. A which has a nested submodule net_b, which itself has two submodules net_c and linear. net_c then has a submodule conv.)

To check whether or not we have the linear submodule, we would call get_submodule("net_b.linear"). To check whether we have the conv submodule, we would call get_submodule("net_b.net_c.conv").

The runtime of get_submodule is bounded by the degree of module nesting in target. A query against named_modules achieves the same result, but it is O(N) in the number of transitive modules. So, for a simple check to see if some submodule exists, get_submodule should always be used.

Parameters:

target (str) – The fully-qualified string name of the submodule to look for. (See above example for how to specify a fully-qualified string.)

Returns:

The submodule referenced by target

Return type:

torch.nn.Module

Raises:

AttributeError – If at any point along the path resulting from the target string the (sub)path resolves to a non-existent attribute name or an object that is not an instance of nn.Module.

half()

Casts all floating point parameters and buffers to half datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

ipu(device=None)

Move all model parameters and buffers to the IPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on IPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

load_state_dict(state_dict, strict=True, assign=False)

Copy parameters and buffers from state_dict into this module and its descendants.

If strict is True, then the keys of state_dict must exactly match the keys returned by this module’s state_dict() function.

Warning

If assign is True the optimizer must be created after the call to load_state_dict unless get_swap_module_params_on_conversion() is True.

Parameters:

• state_dict (dict) – a dict containing parameters and persistent buffers.

• strict (bool, optional) – whether to strictly enforce that the keys in state_dict match the keys returned by this module’s state_dict() function. Default: True

• assign (bool, optional) – When set to False, the properties of the tensors in the current module are preserved whereas setting it to True preserves properties of the Tensors in the state dict. The only exception is the requires_grad field of Parameter for which the value from the module is preserved. Default: False

Returns:

• missing_keys is a list of str containing any keys that are expected

  by this module but missing from the provided state_dict.

• unexpected_keys is a list of str containing the keys that are not

  expected by this module but present in the provided state_dict.

Return type:

NamedTuple with missing_keys and unexpected_keys fields

Note

If a parameter or buffer is registered as None and its corresponding key exists in state_dict, load_state_dict() will raise a RuntimeError.

modules(remove_duplicate=True)

Return an iterator over all modules in the network.

Parameters:

remove_duplicate (bool, default: True) – whether to remove the duplicated module instances in the result or not.

Yields:

Module – a module in the network

Return type:

Iterator[Module]

Note

Duplicate modules are returned only once by default. In the following example, l will be returned only once.

Example:

>>> l = nn.Linear(2, 2)
>>> net = nn.Sequential(l, l)
>>> for idx, m in enumerate(net.modules()):
...     print(idx, '->', m)

0 -> Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
)
1 -> Linear(in_features=2, out_features=2, bias=True)

mtia(device=None)

Move all model parameters and buffers to the MTIA.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on MTIA while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

named_buffers(prefix='', recurse=True, remove_duplicate=True)

Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.

Parameters:

• prefix (str) – prefix to prepend to all buffer names.

• recurse (bool, optional) – if True, then yields buffers of this module and all submodules. Otherwise, yields only buffers that are direct members of this module. Defaults to True.

• remove_duplicate (bool, optional) – whether to remove the duplicated buffers in the result. Defaults to True.

Yields:

(str, torch.Tensor) – Tuple containing the name and buffer

Return type:

Iterator[tuple[str, Tensor]]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, buf in self.named_buffers():
>>>     if name in ['running_var']:
>>>         print(buf.size())

named_children()

Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.

Yields:

(str, Module) – Tuple containing a name and child module

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, module in model.named_children():
>>>     if name in ['conv4', 'conv5']:
>>>         print(module)

Return type:

Iterator[tuple[str, Module]]

named_modules(memo=None, prefix='', remove_duplicate=True)

Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.

Parameters:

• memo (set[Module] | None, default: None) – a memo to store the set of modules already added to the result

• prefix (str, default: '') – a prefix that will be added to the name of the module

• remove_duplicate (bool, default: True) – whether to remove the duplicated module instances in the result or not

Yields:

(str, Module) – Tuple of name and module

Note

Duplicate modules are returned only once. In the following example, l will be returned only once.

Example:

>>> l = nn.Linear(2, 2)
>>> net = nn.Sequential(l, l)
>>> for idx, m in enumerate(net.named_modules()):
...     print(idx, '->', m)

0 -> ('', Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
))
1 -> ('0', Linear(in_features=2, out_features=2, bias=True))

named_parameters(prefix='', recurse=True, remove_duplicate=True)

Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.

Parameters:

• prefix (str) – prefix to prepend to all parameter names.

• recurse (bool) – if True, then yields parameters of this module and all submodules. Otherwise, yields only parameters that are direct members of this module.

• remove_duplicate (bool, optional) – whether to remove the duplicated parameters in the result. Defaults to True.

Yields:

(str, Parameter) – Tuple containing the name and parameter

Return type:

Iterator[tuple[str, Parameter]]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, param in self.named_parameters():
>>>     if name in ['bias']:
>>>         print(param.size())

parameters(recurse=True)

Return an iterator over module parameters.

This is typically passed to an optimizer.

Parameters:

recurse (bool) – if True, then yields parameters of this module and all submodules. Otherwise, yields only parameters that are direct members of this module.

Yields:

Parameter – module parameter

Return type:

Iterator[Parameter]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for param in model.parameters():
>>>     print(type(param), param.size())
<class 'torch.Tensor'> (20L,)
<class 'torch.Tensor'> (20L, 1L, 5L, 5L)

register_backward_hook(hook)

Register a backward hook on the module.

This function is deprecated in favor of register_full_backward_hook() and the behavior of this function will change in future versions.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

Parameters:

hook (Callable[[Module, tuple[Tensor, ...] | Tensor, tuple[Tensor, ...] | Tensor], tuple[Tensor, ...] | Tensor | None])

register_buffer(name, tensor, persistent=True)

Add a buffer to the module.

This is typically used to register a buffer that should not be considered a model parameter. For example, BatchNorm’s running_mean is not a parameter, but is part of the module’s state. Buffers, by default, are persistent and will be saved alongside parameters. This behavior can be changed by setting persistent to False. The only difference between a persistent buffer and a non-persistent buffer is that the latter will not be a part of this module’s state_dict.

Buffers can be accessed as attributes using given names.

Parameters:

• name (str) – name of the buffer. The buffer can be accessed from this module using the given name

• tensor (Tensor or None) – buffer to be registered. If None, then operations that run on buffers, such as cuda, are ignored. If None, the buffer is not included in the module’s state_dict.

• persistent (bool) – whether the buffer is part of this module’s state_dict.

Return type:

None

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> self.register_buffer('running_mean', torch.zeros(num_features))

register_forward_hook(hook, *, prepend=False, with_kwargs=False, always_call=False)

Register a forward hook on the module.

The hook will be called every time after forward() has computed an output.

If with_kwargs is False or not specified, the input contains only the positional arguments given to the module. Keyword arguments won’t be passed to the hooks and only to the forward. The hook can modify the output. It can modify the input inplace but it will not have effect on forward since this is called after forward() is called. The hook should have the following signature:

hook(module, args, output) -> None or modified output

If with_kwargs is True, the forward hook will be passed the kwargs given to the forward function and be expected to return the output possibly modified. The hook should have the following signature:

hook(module, args, kwargs, output) -> None or modified output

Parameters:

• hook (Callable) – The user defined hook to be registered.

• prepend (bool) – If True, the provided hook will be fired before all existing forward hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing forward hooks on this torch.nn.Module. Note that global forward hooks registered with register_module_forward_hook() will fire before all hooks registered by this method. Default: False

• with_kwargs (bool) – If True, the hook will be passed the kwargs given to the forward function. Default: False

• always_call (bool) – If True the hook will be run regardless of whether an exception is raised while calling the Module. Default: False

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_forward_pre_hook(hook, *, prepend=False, with_kwargs=False)

Register a forward pre-hook on the module.

The hook will be called every time before forward() is invoked.

If with_kwargs is false or not specified, the input contains only the positional arguments given to the module. Keyword arguments won’t be passed to the hooks and only to the forward. The hook can modify the input. User can either return a tuple or a single modified value in the hook. We will wrap the value into a tuple if a single value is returned (unless that value is already a tuple). The hook should have the following signature:

hook(module, args) -> None or modified input

If with_kwargs is true, the forward pre-hook will be passed the kwargs given to the forward function. And if the hook modifies the input, both the args and kwargs should be returned. The hook should have the following signature:

hook(module, args, kwargs) -> None or a tuple of modified input and kwargs

Parameters:

• hook (Callable) – The user defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing forward_pre hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing forward_pre hooks on this torch.nn.Module. Note that global forward_pre hooks registered with register_module_forward_pre_hook() will fire before all hooks registered by this method. Default: False

• with_kwargs (bool) – If true, the hook will be passed the kwargs given to the forward function. Default: False

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_full_backward_hook(hook, prepend=False)

Register a backward hook on the module.

The hook will be called every time the gradients with respect to a module are computed, and its firing rules are as follows:

1. Ordinarily, the hook fires when the gradients are computed with respect to the module inputs.

2. If none of the module inputs require gradients, the hook will fire when the gradients are computed with respect to module outputs.

3. If none of the module outputs require gradients, then the hooks will not fire.

The hook should have the following signature:

hook(module, grad_input, grad_output) -> tuple(Tensor) or None

The grad_input and grad_output are tuples that contain the gradients with respect to the inputs and outputs respectively. The hook should not modify its arguments, but it can optionally return a new gradient with respect to the input that will be used in place of grad_input in subsequent computations. grad_input will only correspond to the inputs given as positional arguments and all kwarg arguments are ignored. Entries in grad_input and grad_output will be None for all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will receive a view of each Tensor passed to the Module. Similarly the caller will receive a view of each Tensor returned by the Module’s forward function.

Warning

Modifying inputs or outputs inplace is not allowed when using backward hooks and will raise an error.

Parameters:

• hook (Callable) – The user-defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing backward hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing backward hooks on this torch.nn.Module. Note that global backward hooks registered with register_module_full_backward_hook() will fire before all hooks registered by this method.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_full_backward_pre_hook(hook, prepend=False)

Register a backward pre-hook on the module.

The hook will be called every time the gradients for the module are computed. The hook should have the following signature:

hook(module, grad_output) -> tuple[Tensor, ...], Tensor or None

The grad_output is a tuple. The hook should not modify its arguments, but it can optionally return a new gradient with respect to the output that will be used in place of grad_output in subsequent computations. Entries in grad_output will be None for all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will receive a view of each Tensor passed to the Module. Similarly the caller will receive a view of each Tensor returned by the Module’s forward function.

Warning

Modifying inputs inplace is not allowed when using backward hooks and will raise an error.

Parameters:

• hook (Callable) – The user-defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing backward_pre hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing backward_pre hooks on this torch.nn.Module. Note that global backward_pre hooks registered with register_module_full_backward_pre_hook() will fire before all hooks registered by this method.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_load_state_dict_post_hook(hook)

Register a post-hook to be run after module’s load_state_dict() is called.

It should have the following signature::

hook(module, incompatible_keys) -> None

The module argument is the current module that this hook is registered on, and the incompatible_keys argument is a NamedTuple consisting of attributes missing_keys and unexpected_keys. missing_keys is a list of str containing the missing keys and unexpected_keys is a list of str containing the unexpected keys.

The given incompatible_keys can be modified inplace if needed.

Note that the checks performed when calling load_state_dict() with strict=True are affected by modifications the hook makes to missing_keys or unexpected_keys, as expected. Additions to either set of keys will result in an error being thrown when strict=True, and clearing out both missing and unexpected keys will avoid an error.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_load_state_dict_pre_hook(hook)

Register a pre-hook to be run before module’s load_state_dict() is called.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs) -> None # noqa: B950

Parameters:

hook (Callable) – Callable hook that will be invoked before loading the state dict.

register_module(name, module)

Alias for add_module().

Parameters:

• name (str)

• module (Optional[Module])

Return type:

None

register_parameter(name, param)

Add a parameter to the module.

The parameter can be accessed as an attribute using given name.

Parameters:

• name (str) – name of the parameter. The parameter can be accessed from this module using the given name

• param (Parameter or None) – parameter to be added to the module. If None, then operations that run on parameters, such as cuda, are ignored. If None, the parameter is not included in the module’s state_dict.

Return type:

None

register_state_dict_post_hook(hook)

Register a post-hook for the state_dict() method.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata) -> None

The registered hooks can modify the state_dict inplace.

register_state_dict_pre_hook(hook)

Register a pre-hook for the state_dict() method.

It should have the following signature::

hook(module, prefix, keep_vars) -> None

The registered hooks can be used to perform pre-processing before the state_dict call is made.

requires_grad_(requires_grad=True)

Change if autograd should record operations on parameters in this module.

This method sets the parameters’ requires_grad attributes in-place.

This method is helpful for freezing part of the module for finetuning or training parts of a model individually (e.g., GAN training).

See Locally disabling gradient computation for a comparison between .requires_grad_() and several similar mechanisms that may be confused with it.

Parameters:

requires_grad (bool) – whether autograd should record operations on parameters in this module. Default: True.

Returns:

self

Return type:

Module

set_extra_state(state)

Set extra state contained in the loaded state_dict.

This function is called from load_state_dict() to handle any extra state found within the state_dict. Implement this function and a corresponding get_extra_state() for your module if you need to store extra state within its state_dict.

Parameters:

state (dict) – Extra state from the state_dict

Return type:

None

set_submodule(target, module, strict=False)

Set the submodule given by target if it exists, otherwise throw an error.

Note

If strict is set to False (default), the method will replace an existing submodule or create a new submodule if the parent module exists. If strict is set to True, the method will only attempt to replace an existing submodule and throw an error if the submodule does not exist.

For example, let’s say you have an nn.Module A that looks like this:

A(
    (net_b): Module(
        (net_c): Module(
            (conv): Conv2d(3, 3, 3)
        )
        (linear): Linear(3, 3)
    )
)

(The diagram shows an nn.Module A. A has a nested submodule net_b, which itself has two submodules net_c and linear. net_c then has a submodule conv.)

To override the Conv2d with a new submodule Linear, you could call set_submodule("net_b.net_c.conv", nn.Linear(1, 1)) where strict could be True or False

To add a new submodule Conv2d to the existing net_b module, you would call set_submodule("net_b.conv", nn.Conv2d(1, 1, 1)).

In the above if you set strict=True and call set_submodule("net_b.conv", nn.Conv2d(1, 1, 1), strict=True), an AttributeError will be raised because net_b does not have a submodule named conv.

Parameters:

• target (str) – The fully-qualified string name of the submodule to look for. (See above example for how to specify a fully-qualified string.)

• module (Module) – The module to set the submodule to.

• strict (bool, default: False) – If False, the method will replace an existing submodule or create a new submodule if the parent module exists. If True, the method will only attempt to replace an existing submodule and throw an error if the submodule doesn’t already exist.

Raises:

• ValueError – If the target string is empty or if module is not an instance of nn.Module.

• AttributeError – If at any point along the path resulting from the target string the (sub)path resolves to a non-existent attribute name or an object that is not an instance of nn.Module.

Return type:

None

share_memory()

See torch.Tensor.share_memory_().

Return type:

Self

state_dict(*args, destination=None, prefix='', keep_vars=False)

Return a dictionary containing references to the whole state of the module.

Both parameters and persistent buffers (e.g. running averages) are included. Keys are corresponding parameter and buffer names. Parameters and buffers set to None are not included.

Note

The returned object is a shallow copy. It contains references to the module’s parameters and buffers.

Warning

Currently state_dict() also accepts positional arguments for destination, prefix and keep_vars in order. However, this is being deprecated and keyword arguments will be enforced in future releases.

Warning

Please avoid the use of argument destination as it is not designed for end-users.

Parameters:

• destination (dict, optional) – If provided, the state of module will be updated into the dict and the same object is returned. Otherwise, an OrderedDict will be created and returned. Default: None.

• prefix (str, optional) – a prefix added to parameter and buffer names to compose the keys in state_dict. Default: ''.

• keep_vars (bool, optional) – by default the Tensor s returned in the state dict are detached from autograd. If it’s set to True, detaching will not be performed. Default: False.

Returns:

a dictionary containing a whole state of the module

Return type:

dict

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> module.state_dict().keys()
['bias', 'weight']

to(*args, **kwargs)

Move and/or cast the parameters and buffers.

This can be called as

to(device=None, dtype=None, non_blocking=False)

to(dtype, non_blocking=False)

to(tensor, non_blocking=False)

to(memory_format=torch.channels_last)

Its signature is similar to torch.Tensor.to(), but only accepts floating point or complex dtypes. In addition, this method will only cast the floating point or complex parameters and buffers to dtype (if given). The integral parameters and buffers will be moved device, if that is given, but with dtypes unchanged. When non_blocking is set, it tries to convert/move asynchronously with respect to the host if possible, e.g., moving CPU Tensors with pinned memory to CUDA devices.

See below for examples.

Note

This method modifies the module in-place.

Parameters:

• device (torch.device) – the desired device of the parameters and buffers in this module

• dtype (torch.dtype) – the desired floating point or complex dtype of the parameters and buffers in this module

• tensor (torch.Tensor) – Tensor whose dtype and device are the desired dtype and device for all parameters and buffers in this module

• memory_format (torch.memory_format) – the desired memory format for 4D parameters and buffers in this module (keyword only argument)

Returns:

self

Return type:

Module

Examples:

>>> # xdoctest: +IGNORE_WANT("non-deterministic")
>>> linear = nn.Linear(2, 2)
>>> linear.weight
Parameter containing:
tensor([[ 0.1913, -0.3420],
        [-0.5113, -0.2325]])
>>> linear.to(torch.double)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1913, -0.3420],
        [-0.5113, -0.2325]], dtype=torch.float64)
>>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA1)
>>> gpu1 = torch.device("cuda:1")
>>> linear.to(gpu1, dtype=torch.half, non_blocking=True)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1914, -0.3420],
        [-0.5112, -0.2324]], dtype=torch.float16, device='cuda:1')
>>> cpu = torch.device("cpu")
>>> linear.to(cpu)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1914, -0.3420],
        [-0.5112, -0.2324]], dtype=torch.float16)

>>> linear = nn.Linear(2, 2, bias=None).to(torch.cdouble)
>>> linear.weight
Parameter containing:
tensor([[ 0.3741+0.j,  0.2382+0.j],
        [ 0.5593+0.j, -0.4443+0.j]], dtype=torch.complex128)
>>> linear(torch.ones(3, 2, dtype=torch.cdouble))
tensor([[0.6122+0.j, 0.1150+0.j],
        [0.6122+0.j, 0.1150+0.j],
        [0.6122+0.j, 0.1150+0.j]], dtype=torch.complex128)

to_empty(*, device, recurse=True)

Move the parameters and buffers to the specified device without copying storage.

Parameters:

• device (torch.device) – The desired device of the parameters and buffers in this module.

• recurse (bool) – Whether parameters and buffers of submodules should be recursively moved to the specified device.

Returns:

self

Return type:

Module

train(mode=True)

Set the module in training mode.

This has an effect only on certain modules. See the documentation of particular modules for details of their behaviors in training/evaluation mode, i.e., whether they are affected, e.g. Dropout, BatchNorm, etc.

Parameters:

mode (bool) – whether to set training mode (True) or evaluation mode (False). Default: True.

Returns:

self

Return type:

Module

type(dst_type)

Casts all parameters and buffers to dst_type.

Note

This method modifies the module in-place.

Parameters:

dst_type (type or string) – the desired type

Returns:

self

Return type:

Module

xpu(device=None)

Move all model parameters and buffers to the XPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing optimizer if the module will live on XPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

zero_grad(set_to_none=True)

Reset gradients of all model parameters.

See similar function under torch.optim.Optimizer for more context.

Parameters:

set_to_none (bool) – instead of setting to zero, set the grads to None. See torch.optim.Optimizer.zero_grad() for details.

Return type:

None

training: bool

class medlatents.diffusion.ScoreEntropyLoss(vocab_size, entropy_weight=0.1, temperature=1.0)

Bases: torch.nn.modules.module.Module

Pure score entropy loss for discrete diffusion.

Directly minimizes the entropy of the predicted score distribution, encouraging the model to be confident in its predictions.

This is a simpler alternative to full SEDD that still captures the key benefits of score-based modeling.

Parameters:

• vocab_size (int) – Number of discrete classes

• entropy_weight (float, default: 0.1) – Weight for entropy regularization

• temperature (float, default: 1.0) – Temperature for softmax

__init__(vocab_size, entropy_weight=0.1, temperature=1.0)

Initialize internal Module state, shared by both nn.Module and ScriptModule.

Parameters:

• vocab_size (int)

• entropy_weight (float, default: 0.1)

• temperature (float, default: 1.0)

forward(logits, targets, t=None)

Compute score entropy loss.

Parameters:

• logits (Tensor) – [batch, seq_len, vocab_size] predicted logits

• targets (Tensor) – [batch, seq_len] target tokens

• t (Tensor | None, default: None) – [batch] optional time values for weighting

Returns:

Scalar loss metrics: Dict with loss components

Return type:

loss

T_destination = ~T_destination

add_module(name, module)

Add a child module to the current module.

The module can be accessed as an attribute using the given name.

Parameters:

• name (str) – name of the child module. The child module can be accessed from this module using the given name

• module (Module) – child module to be added to the module.

Return type:

None

apply(fn)

Apply fn recursively to every submodule (as returned by .children()) as well as self.

Typical use includes initializing the parameters of a model (see also torch.nn.init).

Parameters:

fn (Module -> None) – function to be applied to each submodule

Returns:

self

Return type:

Module

Example:

>>> @torch.no_grad()
>>> def init_weights(m):
>>>     print(m)
>>>     if type(m) is nn.Linear:
>>>         m.weight.fill_(1.0)
>>>         print(m.weight)
>>> net = nn.Sequential(nn.Linear(2, 2), nn.Linear(2, 2))
>>> net.apply(init_weights)
Linear(in_features=2, out_features=2, bias=True)
Parameter containing:
tensor([[1., 1.],
        [1., 1.]], requires_grad=True)
Linear(in_features=2, out_features=2, bias=True)
Parameter containing:
tensor([[1., 1.],
        [1., 1.]], requires_grad=True)
Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
)

bfloat16()

Casts all floating point parameters and buffers to bfloat16 datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

buffers(recurse=True)

Return an iterator over module buffers.

Parameters:

recurse (bool) – if True, then yields buffers of this module and all submodules. Otherwise, yields only buffers that are direct members of this module.

Yields:

torch.Tensor – module buffer

Return type:

Iterator[Tensor]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for buf in model.buffers():
>>>     print(type(buf), buf.size())
<class 'torch.Tensor'> (20L,)
<class 'torch.Tensor'> (20L, 1L, 5L, 5L)

call_super_init: bool = False

children()

Return an iterator over immediate children modules.

Yields:

Module – a child module

Return type:

Iterator[Module]

compile(*args, **kwargs)

Compile this Module’s forward using torch.compile().

This Module’s __call__ method is compiled and all arguments are passed as-is to torch.compile().

See torch.compile() for details on the arguments for this function.

Return type:

None

cpu()

Move all model parameters and buffers to the CPU.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

cuda(device=None)

Move all model parameters and buffers to the GPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on GPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

double()

Casts all floating point parameters and buffers to double datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

dump_patches: bool = False

eval()

Set the module in evaluation mode.

This has an effect only on certain modules. See the documentation of particular modules for details of their behaviors in training/evaluation mode, i.e. whether they are affected, e.g. Dropout, BatchNorm, etc.

This is equivalent with self.train(False).

See Locally disabling gradient computation for a comparison between .eval() and several similar mechanisms that may be confused with it.

Returns:

self

Return type:

Module

extra_repr()

Return the extra representation of the module.

To print customized extra information, you should re-implement this method in your own modules. Both single-line and multi-line strings are acceptable.

Return type:

str

float()

Casts all floating point parameters and buffers to float datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

get_buffer(target)

Return the buffer given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed explanation of this method’s functionality as well as how to correctly specify target.

Parameters:

target (str) – The fully-qualified string name of the buffer to look for. (See get_submodule for how to specify a fully-qualified string.)

Returns:

The buffer referenced by target

Return type:

torch.Tensor

Raises:

AttributeError – If the target string references an invalid path or resolves to something that is not a buffer

get_extra_state()

Return any extra state to include in the module’s state_dict.

Implement this and a corresponding set_extra_state() for your module if you need to store extra state. This function is called when building the module’s state_dict().

Note that extra state should be picklable to ensure working serialization of the state_dict. We only provide backwards compatibility guarantees for serializing Tensors; other objects may break backwards compatibility if their serialized pickled form changes.

Returns:

Any extra state to store in the module’s state_dict

Return type:

object

get_parameter(target)

Return the parameter given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed explanation of this method’s functionality as well as how to correctly specify target.

Parameters:

target (str) – The fully-qualified string name of the Parameter to look for. (See get_submodule for how to specify a fully-qualified string.)

Returns:

The Parameter referenced by target

Return type:

torch.nn.Parameter

Raises:

AttributeError – If the target string references an invalid path or resolves to something that is not an nn.Parameter

get_submodule(target)

Return the submodule given by target if it exists, otherwise throw an error.

For example, let’s say you have an nn.Module A that looks like this:

A(
    (net_b): Module(
        (net_c): Module(
            (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))
        )
        (linear): Linear(in_features=100, out_features=200, bias=True)
    )
)

(The diagram shows an nn.Module A. A which has a nested submodule net_b, which itself has two submodules net_c and linear. net_c then has a submodule conv.)

To check whether or not we have the linear submodule, we would call get_submodule("net_b.linear"). To check whether we have the conv submodule, we would call get_submodule("net_b.net_c.conv").

The runtime of get_submodule is bounded by the degree of module nesting in target. A query against named_modules achieves the same result, but it is O(N) in the number of transitive modules. So, for a simple check to see if some submodule exists, get_submodule should always be used.

Parameters:

target (str) – The fully-qualified string name of the submodule to look for. (See above example for how to specify a fully-qualified string.)

Returns:

The submodule referenced by target

Return type:

torch.nn.Module

Raises:

AttributeError – If at any point along the path resulting from the target string the (sub)path resolves to a non-existent attribute name or an object that is not an instance of nn.Module.

half()

Casts all floating point parameters and buffers to half datatype.

Note

This method modifies the module in-place.

Returns:

self

Return type:

Module

ipu(device=None)

Move all model parameters and buffers to the IPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on IPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

load_state_dict(state_dict, strict=True, assign=False)

Copy parameters and buffers from state_dict into this module and its descendants.

If strict is True, then the keys of state_dict must exactly match the keys returned by this module’s state_dict() function.

Warning

If assign is True the optimizer must be created after the call to load_state_dict unless get_swap_module_params_on_conversion() is True.

Parameters:

• state_dict (dict) – a dict containing parameters and persistent buffers.

• strict (bool, optional) – whether to strictly enforce that the keys in state_dict match the keys returned by this module’s state_dict() function. Default: True

• assign (bool, optional) – When set to False, the properties of the tensors in the current module are preserved whereas setting it to True preserves properties of the Tensors in the state dict. The only exception is the requires_grad field of Parameter for which the value from the module is preserved. Default: False

Returns:

• missing_keys is a list of str containing any keys that are expected

  by this module but missing from the provided state_dict.

• unexpected_keys is a list of str containing the keys that are not

  expected by this module but present in the provided state_dict.

Return type:

NamedTuple with missing_keys and unexpected_keys fields

Note

If a parameter or buffer is registered as None and its corresponding key exists in state_dict, load_state_dict() will raise a RuntimeError.

modules(remove_duplicate=True)

Return an iterator over all modules in the network.

Parameters:

remove_duplicate (bool, default: True) – whether to remove the duplicated module instances in the result or not.

Yields:

Module – a module in the network

Return type:

Iterator[Module]

Note

Duplicate modules are returned only once by default. In the following example, l will be returned only once.

Example:

>>> l = nn.Linear(2, 2)
>>> net = nn.Sequential(l, l)
>>> for idx, m in enumerate(net.modules()):
...     print(idx, '->', m)

0 -> Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
)
1 -> Linear(in_features=2, out_features=2, bias=True)

mtia(device=None)

Move all model parameters and buffers to the MTIA.

This also makes associated parameters and buffers different objects. So it should be called before constructing the optimizer if the module will live on MTIA while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

named_buffers(prefix='', recurse=True, remove_duplicate=True)

Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.

Parameters:

• prefix (str) – prefix to prepend to all buffer names.

• recurse (bool, optional) – if True, then yields buffers of this module and all submodules. Otherwise, yields only buffers that are direct members of this module. Defaults to True.

• remove_duplicate (bool, optional) – whether to remove the duplicated buffers in the result. Defaults to True.

Yields:

(str, torch.Tensor) – Tuple containing the name and buffer

Return type:

Iterator[tuple[str, Tensor]]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, buf in self.named_buffers():
>>>     if name in ['running_var']:
>>>         print(buf.size())

named_children()

Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.

Yields:

(str, Module) – Tuple containing a name and child module

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, module in model.named_children():
>>>     if name in ['conv4', 'conv5']:
>>>         print(module)

Return type:

Iterator[tuple[str, Module]]

named_modules(memo=None, prefix='', remove_duplicate=True)

Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.

Parameters:

• memo (set[Module] | None, default: None) – a memo to store the set of modules already added to the result

• prefix (str, default: '') – a prefix that will be added to the name of the module

• remove_duplicate (bool, default: True) – whether to remove the duplicated module instances in the result or not

Yields:

(str, Module) – Tuple of name and module

Note

Duplicate modules are returned only once. In the following example, l will be returned only once.

Example:

>>> l = nn.Linear(2, 2)
>>> net = nn.Sequential(l, l)
>>> for idx, m in enumerate(net.named_modules()):
...     print(idx, '->', m)

0 -> ('', Sequential(
  (0): Linear(in_features=2, out_features=2, bias=True)
  (1): Linear(in_features=2, out_features=2, bias=True)
))
1 -> ('0', Linear(in_features=2, out_features=2, bias=True))

named_parameters(prefix='', recurse=True, remove_duplicate=True)

Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.

Parameters:

• prefix (str) – prefix to prepend to all parameter names.

• recurse (bool) – if True, then yields parameters of this module and all submodules. Otherwise, yields only parameters that are direct members of this module.

• remove_duplicate (bool, optional) – whether to remove the duplicated parameters in the result. Defaults to True.

Yields:

(str, Parameter) – Tuple containing the name and parameter

Return type:

Iterator[tuple[str, Parameter]]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for name, param in self.named_parameters():
>>>     if name in ['bias']:
>>>         print(param.size())

parameters(recurse=True)

Return an iterator over module parameters.

This is typically passed to an optimizer.

Parameters:

recurse (bool) – if True, then yields parameters of this module and all submodules. Otherwise, yields only parameters that are direct members of this module.

Yields:

Parameter – module parameter

Return type:

Iterator[Parameter]

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> for param in model.parameters():
>>>     print(type(param), param.size())
<class 'torch.Tensor'> (20L,)
<class 'torch.Tensor'> (20L, 1L, 5L, 5L)

register_backward_hook(hook)

Register a backward hook on the module.

This function is deprecated in favor of register_full_backward_hook() and the behavior of this function will change in future versions.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

Parameters:

hook (Callable[[Module, tuple[Tensor, ...] | Tensor, tuple[Tensor, ...] | Tensor], tuple[Tensor, ...] | Tensor | None])

register_buffer(name, tensor, persistent=True)

Add a buffer to the module.

This is typically used to register a buffer that should not be considered a model parameter. For example, BatchNorm’s running_mean is not a parameter, but is part of the module’s state. Buffers, by default, are persistent and will be saved alongside parameters. This behavior can be changed by setting persistent to False. The only difference between a persistent buffer and a non-persistent buffer is that the latter will not be a part of this module’s state_dict.

Buffers can be accessed as attributes using given names.

Parameters:

• name (str) – name of the buffer. The buffer can be accessed from this module using the given name

• tensor (Tensor or None) – buffer to be registered. If None, then operations that run on buffers, such as cuda, are ignored. If None, the buffer is not included in the module’s state_dict.

• persistent (bool) – whether the buffer is part of this module’s state_dict.

Return type:

None

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> self.register_buffer('running_mean', torch.zeros(num_features))

register_forward_hook(hook, *, prepend=False, with_kwargs=False, always_call=False)

Register a forward hook on the module.

The hook will be called every time after forward() has computed an output.

If with_kwargs is False or not specified, the input contains only the positional arguments given to the module. Keyword arguments won’t be passed to the hooks and only to the forward. The hook can modify the output. It can modify the input inplace but it will not have effect on forward since this is called after forward() is called. The hook should have the following signature:

hook(module, args, output) -> None or modified output

If with_kwargs is True, the forward hook will be passed the kwargs given to the forward function and be expected to return the output possibly modified. The hook should have the following signature:

hook(module, args, kwargs, output) -> None or modified output

Parameters:

• hook (Callable) – The user defined hook to be registered.

• prepend (bool) – If True, the provided hook will be fired before all existing forward hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing forward hooks on this torch.nn.Module. Note that global forward hooks registered with register_module_forward_hook() will fire before all hooks registered by this method. Default: False

• with_kwargs (bool) – If True, the hook will be passed the kwargs given to the forward function. Default: False

• always_call (bool) – If True the hook will be run regardless of whether an exception is raised while calling the Module. Default: False

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_forward_pre_hook(hook, *, prepend=False, with_kwargs=False)

Register a forward pre-hook on the module.

The hook will be called every time before forward() is invoked.

If with_kwargs is false or not specified, the input contains only the positional arguments given to the module. Keyword arguments won’t be passed to the hooks and only to the forward. The hook can modify the input. User can either return a tuple or a single modified value in the hook. We will wrap the value into a tuple if a single value is returned (unless that value is already a tuple). The hook should have the following signature:

hook(module, args) -> None or modified input

If with_kwargs is true, the forward pre-hook will be passed the kwargs given to the forward function. And if the hook modifies the input, both the args and kwargs should be returned. The hook should have the following signature:

hook(module, args, kwargs) -> None or a tuple of modified input and kwargs

Parameters:

• hook (Callable) – The user defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing forward_pre hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing forward_pre hooks on this torch.nn.Module. Note that global forward_pre hooks registered with register_module_forward_pre_hook() will fire before all hooks registered by this method. Default: False

• with_kwargs (bool) – If true, the hook will be passed the kwargs given to the forward function. Default: False

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_full_backward_hook(hook, prepend=False)

Register a backward hook on the module.

The hook will be called every time the gradients with respect to a module are computed, and its firing rules are as follows:

1. Ordinarily, the hook fires when the gradients are computed with respect to the module inputs.

2. If none of the module inputs require gradients, the hook will fire when the gradients are computed with respect to module outputs.

3. If none of the module outputs require gradients, then the hooks will not fire.

The hook should have the following signature:

hook(module, grad_input, grad_output) -> tuple(Tensor) or None

The grad_input and grad_output are tuples that contain the gradients with respect to the inputs and outputs respectively. The hook should not modify its arguments, but it can optionally return a new gradient with respect to the input that will be used in place of grad_input in subsequent computations. grad_input will only correspond to the inputs given as positional arguments and all kwarg arguments are ignored. Entries in grad_input and grad_output will be None for all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will receive a view of each Tensor passed to the Module. Similarly the caller will receive a view of each Tensor returned by the Module’s forward function.

Warning

Modifying inputs or outputs inplace is not allowed when using backward hooks and will raise an error.

Parameters:

• hook (Callable) – The user-defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing backward hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing backward hooks on this torch.nn.Module. Note that global backward hooks registered with register_module_full_backward_hook() will fire before all hooks registered by this method.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_full_backward_pre_hook(hook, prepend=False)

Register a backward pre-hook on the module.

The hook will be called every time the gradients for the module are computed. The hook should have the following signature:

hook(module, grad_output) -> tuple[Tensor, ...], Tensor or None

The grad_output is a tuple. The hook should not modify its arguments, but it can optionally return a new gradient with respect to the output that will be used in place of grad_output in subsequent computations. Entries in grad_output will be None for all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will receive a view of each Tensor passed to the Module. Similarly the caller will receive a view of each Tensor returned by the Module’s forward function.

Warning

Modifying inputs inplace is not allowed when using backward hooks and will raise an error.

Parameters:

• hook (Callable) – The user-defined hook to be registered.

• prepend (bool) – If true, the provided hook will be fired before all existing backward_pre hooks on this torch.nn.Module. Otherwise, the provided hook will be fired after all existing backward_pre hooks on this torch.nn.Module. Note that global backward_pre hooks registered with register_module_full_backward_pre_hook() will fire before all hooks registered by this method.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_load_state_dict_post_hook(hook)

Register a post-hook to be run after module’s load_state_dict() is called.

It should have the following signature::

hook(module, incompatible_keys) -> None

The module argument is the current module that this hook is registered on, and the incompatible_keys argument is a NamedTuple consisting of attributes missing_keys and unexpected_keys. missing_keys is a list of str containing the missing keys and unexpected_keys is a list of str containing the unexpected keys.

The given incompatible_keys can be modified inplace if needed.

Note that the checks performed when calling load_state_dict() with strict=True are affected by modifications the hook makes to missing_keys or unexpected_keys, as expected. Additions to either set of keys will result in an error being thrown when strict=True, and clearing out both missing and unexpected keys will avoid an error.

Returns:

a handle that can be used to remove the added hook by calling handle.remove()

Return type:

torch.utils.hooks.RemovableHandle

register_load_state_dict_pre_hook(hook)

Register a pre-hook to be run before module’s load_state_dict() is called.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs) -> None # noqa: B950

Parameters:

hook (Callable) – Callable hook that will be invoked before loading the state dict.

register_module(name, module)

Alias for add_module().

Parameters:

• name (str)

• module (Optional[Module])

Return type:

None

register_parameter(name, param)

Add a parameter to the module.

The parameter can be accessed as an attribute using given name.

Parameters:

• name (str) – name of the parameter. The parameter can be accessed from this module using the given name

• param (Parameter or None) – parameter to be added to the module. If None, then operations that run on parameters, such as cuda, are ignored. If None, the parameter is not included in the module’s state_dict.

Return type:

None

register_state_dict_post_hook(hook)

Register a post-hook for the state_dict() method.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata) -> None

The registered hooks can modify the state_dict inplace.

register_state_dict_pre_hook(hook)

Register a pre-hook for the state_dict() method.

It should have the following signature::

hook(module, prefix, keep_vars) -> None

The registered hooks can be used to perform pre-processing before the state_dict call is made.

requires_grad_(requires_grad=True)

Change if autograd should record operations on parameters in this module.

This method sets the parameters’ requires_grad attributes in-place.

This method is helpful for freezing part of the module for finetuning or training parts of a model individually (e.g., GAN training).

See Locally disabling gradient computation for a comparison between .requires_grad_() and several similar mechanisms that may be confused with it.

Parameters:

requires_grad (bool) – whether autograd should record operations on parameters in this module. Default: True.

Returns:

self

Return type:

Module

set_extra_state(state)

Set extra state contained in the loaded state_dict.

This function is called from load_state_dict() to handle any extra state found within the state_dict. Implement this function and a corresponding get_extra_state() for your module if you need to store extra state within its state_dict.

Parameters:

state (dict) – Extra state from the state_dict

Return type:

None

set_submodule(target, module, strict=False)

Set the submodule given by target if it exists, otherwise throw an error.

Note

If strict is set to False (default), the method will replace an existing submodule or create a new submodule if the parent module exists. If strict is set to True, the method will only attempt to replace an existing submodule and throw an error if the submodule does not exist.

For example, let’s say you have an nn.Module A that looks like this:

A(
    (net_b): Module(
        (net_c): Module(
            (conv): Conv2d(3, 3, 3)
        )
        (linear): Linear(3, 3)
    )
)

(The diagram shows an nn.Module A. A has a nested submodule net_b, which itself has two submodules net_c and linear. net_c then has a submodule conv.)

To override the Conv2d with a new submodule Linear, you could call set_submodule("net_b.net_c.conv", nn.Linear(1, 1)) where strict could be True or False

To add a new submodule Conv2d to the existing net_b module, you would call set_submodule("net_b.conv", nn.Conv2d(1, 1, 1)).

In the above if you set strict=True and call set_submodule("net_b.conv", nn.Conv2d(1, 1, 1), strict=True), an AttributeError will be raised because net_b does not have a submodule named conv.

Parameters:

• target (str) – The fully-qualified string name of the submodule to look for. (See above example for how to specify a fully-qualified string.)

• module (Module) – The module to set the submodule to.

• strict (bool, default: False) – If False, the method will replace an existing submodule or create a new submodule if the parent module exists. If True, the method will only attempt to replace an existing submodule and throw an error if the submodule doesn’t already exist.

Raises:

• ValueError – If the target string is empty or if module is not an instance of nn.Module.

• AttributeError – If at any point along the path resulting from the target string the (sub)path resolves to a non-existent attribute name or an object that is not an instance of nn.Module.

Return type:

None

share_memory()

See torch.Tensor.share_memory_().

Return type:

Self

state_dict(*args, destination=None, prefix='', keep_vars=False)

Return a dictionary containing references to the whole state of the module.

Both parameters and persistent buffers (e.g. running averages) are included. Keys are corresponding parameter and buffer names. Parameters and buffers set to None are not included.

Note

The returned object is a shallow copy. It contains references to the module’s parameters and buffers.

Warning

Currently state_dict() also accepts positional arguments for destination, prefix and keep_vars in order. However, this is being deprecated and keyword arguments will be enforced in future releases.

Warning

Please avoid the use of argument destination as it is not designed for end-users.

Parameters:

• destination (dict, optional) – If provided, the state of module will be updated into the dict and the same object is returned. Otherwise, an OrderedDict will be created and returned. Default: None.

• prefix (str, optional) – a prefix added to parameter and buffer names to compose the keys in state_dict. Default: ''.

• keep_vars (bool, optional) – by default the Tensor s returned in the state dict are detached from autograd. If it’s set to True, detaching will not be performed. Default: False.

Returns:

a dictionary containing a whole state of the module

Return type:

dict

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> module.state_dict().keys()
['bias', 'weight']

to(*args, **kwargs)

Move and/or cast the parameters and buffers.

This can be called as

to(device=None, dtype=None, non_blocking=False)

to(dtype, non_blocking=False)

to(tensor, non_blocking=False)

to(memory_format=torch.channels_last)

Its signature is similar to torch.Tensor.to(), but only accepts floating point or complex dtypes. In addition, this method will only cast the floating point or complex parameters and buffers to dtype (if given). The integral parameters and buffers will be moved device, if that is given, but with dtypes unchanged. When non_blocking is set, it tries to convert/move asynchronously with respect to the host if possible, e.g., moving CPU Tensors with pinned memory to CUDA devices.

See below for examples.

Note

This method modifies the module in-place.

Parameters:

• device (torch.device) – the desired device of the parameters and buffers in this module

• dtype (torch.dtype) – the desired floating point or complex dtype of the parameters and buffers in this module

• tensor (torch.Tensor) – Tensor whose dtype and device are the desired dtype and device for all parameters and buffers in this module

• memory_format (torch.memory_format) – the desired memory format for 4D parameters and buffers in this module (keyword only argument)

Returns:

self

Return type:

Module

Examples:

>>> # xdoctest: +IGNORE_WANT("non-deterministic")
>>> linear = nn.Linear(2, 2)
>>> linear.weight
Parameter containing:
tensor([[ 0.1913, -0.3420],
        [-0.5113, -0.2325]])
>>> linear.to(torch.double)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1913, -0.3420],
        [-0.5113, -0.2325]], dtype=torch.float64)
>>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA1)
>>> gpu1 = torch.device("cuda:1")
>>> linear.to(gpu1, dtype=torch.half, non_blocking=True)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1914, -0.3420],
        [-0.5112, -0.2324]], dtype=torch.float16, device='cuda:1')
>>> cpu = torch.device("cpu")
>>> linear.to(cpu)
Linear(in_features=2, out_features=2, bias=True)
>>> linear.weight
Parameter containing:
tensor([[ 0.1914, -0.3420],
        [-0.5112, -0.2324]], dtype=torch.float16)

>>> linear = nn.Linear(2, 2, bias=None).to(torch.cdouble)
>>> linear.weight
Parameter containing:
tensor([[ 0.3741+0.j,  0.2382+0.j],
        [ 0.5593+0.j, -0.4443+0.j]], dtype=torch.complex128)
>>> linear(torch.ones(3, 2, dtype=torch.cdouble))
tensor([[0.6122+0.j, 0.1150+0.j],
        [0.6122+0.j, 0.1150+0.j],
        [0.6122+0.j, 0.1150+0.j]], dtype=torch.complex128)

to_empty(*, device, recurse=True)

Move the parameters and buffers to the specified device without copying storage.

Parameters:

• device (torch.device) – The desired device of the parameters and buffers in this module.

• recurse (bool) – Whether parameters and buffers of submodules should be recursively moved to the specified device.

Returns:

self

Return type:

Module

train(mode=True)

Set the module in training mode.

This has an effect only on certain modules. See the documentation of particular modules for details of their behaviors in training/evaluation mode, i.e., whether they are affected, e.g. Dropout, BatchNorm, etc.

Parameters:

mode (bool) – whether to set training mode (True) or evaluation mode (False). Default: True.

Returns:

self

Return type:

Module

type(dst_type)

Casts all parameters and buffers to dst_type.

Note

This method modifies the module in-place.

Parameters:

dst_type (type or string) – the desired type

Returns:

self

Return type:

Module

xpu(device=None)

Move all model parameters and buffers to the XPU.

This also makes associated parameters and buffers different objects. So it should be called before constructing optimizer if the module will live on XPU while being optimized.

Note

This method modifies the module in-place.

Parameters:

device (int, optional) – if specified, all parameters will be copied to that device

Returns:

self

Return type:

Module

zero_grad(set_to_none=True)

Reset gradients of all model parameters.

See similar function under torch.optim.Optimizer for more context.

Parameters:

set_to_none (bool) – instead of setting to zero, set the grads to None. See torch.optim.Optimizer.zero_grad() for details.

Return type:

None

training: bool

Schedules

Beta schedules and transition matrices for diffusion.

medlatents.diffusion.get_beta_schedule(schedule_type, num_timesteps, beta_start=0.0001, beta_end=0.02, cosine_s=0.008)

Get noise schedule (beta) for diffusion process.

Parameters:

• schedule_type (Literal['linear', 'cosine', 'quadratic', 'sigmoid']) – Type of schedule

• num_timesteps (int) – Number of diffusion timesteps

• beta_start (float, default: 0.0001) – Starting beta value (for linear/sigmoid)

• beta_end (float, default: 0.02) – Ending beta value (for linear/sigmoid)

• cosine_s (float, default: 0.008) – Offset for cosine schedule

Returns:

[num_timesteps] beta values

Return type:

betas

medlatents.diffusion.get_absorbing_transition_mat(num_classes, beta, absorbing_state=None)

Get transition matrix for absorbing diffusion (D3PM variant).

The absorbing state is a special “mask” state that tokens transition to.

Parameters:

• num_classes (int) – Number of classes (vocabulary size)

• beta (float) – Noise level at this timestep

• absorbing_state (int | None, default: None) – Index of absorbing state (default: num_classes)

Returns:

[num_classes+1, num_classes+1] transition matrix

Return type:

Q_t

medlatents.diffusion.get_uniform_transition_mat(num_classes, beta)

Get transition matrix for uniform diffusion (original D3PM).

Tokens transition uniformly to all other classes.

Parameters:

• num_classes (int) – Number of classes

• beta (float) – Noise level at this timestep

Returns:

[num_classes, num_classes] transition matrix

Return type:

Q_t

medlatents.diffusion.get_discretized_gaussian_transition_mat(num_classes, beta, sigma=1.0)

Get transition matrix based on discretized Gaussian kernel.

This creates a “blurring” effect similar to continuous diffusion.

Parameters:

• num_classes (int) – Number of classes

• beta (float) – Noise level at this timestep

• sigma (float, default: 1.0) – Gaussian standard deviation

Returns:

[num_classes, num_classes] transition matrix

Return type:

Q_t

medlatents.diffusion.compute_transition_matrices(betas, num_classes, transition_type='absorbing')

Compute transition matrices for all timesteps.

Parameters:

• betas (Tensor) – [num_timesteps] beta schedule

• num_classes (int) – Number of classes

• transition_type (Literal['absorbing', 'uniform', 'gaussian'], default: 'absorbing') – Type of transition matrix

Returns:

[num_timesteps, num_classes, num_classes] single-step transitions Q_bar_t: [num_timesteps, num_classes, num_classes] cumulative transitions Q_bar_t_minus_1: [num_timesteps, num_classes, num_classes] cumulative for t-1

Return type:

Q_t

Zero-Terminal SNR

Improved training with zero terminal SNR (WACV 2024).

medlatents.diffusion.enforce_zero_terminal_snr(betas)

Enforce zero terminal SNR by rescaling alphas_cumprod.

Standard schedules have non-zero SNR at the final timestep, causing train/inference mismatch. This correction ensures the schedule ends at pure noise (SNR=0).

Reference: “Common Diffusion Noise Schedules and Sample Steps are Flawed” (WACV 2024)

Parameters:

betas (Tensor) – Original beta schedule [timesteps]

Return type:

Tensor

Returns:

Corrected betas with zero terminal SNR

medlatents.diffusion.compute_snr(alphas_cumprod)

Compute Signal-to-Noise Ratio from alphas_cumprod.

SNR(t) = alpha_bar_t / (1 - alpha_bar_t)

Parameters:

alphas_cumprod (Tensor) – Cumulative product of alphas [timesteps]

Return type:

Tensor

Returns:

SNR values [timesteps]

medlatents.diffusion.compute_log_snr(alphas_cumprod)

Compute log Signal-to-Noise Ratio.

log_SNR(t) = log(alpha_bar_t) - log(1 - alpha_bar_t)

Useful for importance sampling (sample more at log_SNR ≈ 0).

Parameters:

alphas_cumprod (Tensor)

Return type:

Tensor

medlatents.diffusion.min_snr_weighting(snr, gamma=5.0)

Min-SNR weighting for loss.

Reweights the loss to emphasize timesteps with lower SNR, which are harder to denoise.

Reference: “Efficient Diffusion Training via Min-SNR Weighting Strategy”

Parameters:

• snr (Tensor) – SNR values at each timestep

• gamma (float, default: 5.0) – Min-SNR gamma value (higher = more uniform weighting)

Return type:

Tensor

Returns:

Weight for each timestep

medlatents.diffusion.get_beta_schedule_with_zero_snr(schedule_type, num_timesteps, beta_start=0.0001, beta_end=0.02, cosine_s=0.008)

Get noise schedule with Zero-Terminal SNR correction.

This ensures the final timestep has SNR=0 (pure noise), fixing the train/inference mismatch in standard schedules.

Parameters:

• schedule_type (Literal['linear', 'cosine', 'quadratic', 'sigmoid']) – Type of schedule

• num_timesteps (int) – Number of diffusion timesteps

• beta_start (float, default: 0.0001) – Starting beta value (for linear/sigmoid)

• beta_end (float, default: 0.02) – Ending beta value (for linear/sigmoid)

• cosine_s (float, default: 0.008) – Offset for cosine schedule

Returns:

[num_timesteps] corrected beta values

Return type:

betas

medlatents.diffusion.sample_timesteps_log_snr_importance(batch_size, device, log_snr, concentration=1.0)

Sample timesteps with importance sampling around log_SNR = 0.

The log_SNR = 0 point (where signal and noise are equal) is often the most important for learning. This samples more timesteps there.

Parameters:

• batch_size (int) – Number of timesteps to sample

• device (device) – Device for output tensor

• log_snr (Tensor) – Pre-computed log-SNR values [num_timesteps]

• concentration (float, default: 1.0) – Higher = more concentration around log_SNR=0

Return type:

Tensor

Returns:

Sampled timestep indices [batch_size]