Module `neuralops.tests.test_replace_comprehensive`

Functions

def test_build_attention()

Expand source code

def test_build_attention():
    attn = build_attention("dummy_test_replace_attn", 64, 4)
    assert isinstance(attn, DummyAttention)
    assert attn.hidden_size == 64

def test_replace_bert_attention()

Expand source code

def test_replace_bert_attention():
    model = DummyBertModel()
    replace_attention_in_model(model, "dummy_test_replace_attn")
    assert isinstance(model.encoder.layer[0].attention.self, DummyAttention)

def test_replace_gpt_attention()

Expand source code

def test_replace_gpt_attention():
    model = DummyGPT2Model()
    replace_attention_in_model(model, "dummy_test_replace_attn")
    assert isinstance(model.transformer.h[0].attn, DummyAttention)

def test_replace_unsupported_model_type()

Expand source code

def test_replace_unsupported_model_type():
    model = nn.Module()
    model.config = SimpleNamespace(model_type="unknown_xyz")
    with pytest.raises(ValueError, match="Unsupported model type"):
        replace_attention_in_model(model, "dummy_test_replace_attn")

Classes

class DummyAttention (hidden_size=64, num_heads=4, **kwargs)

Expand source code

class DummyAttention(nn.Module):
    def __init__(self, hidden_size=64, num_heads=4, **kwargs):
        super().__init__()
        self.hidden_size = hidden_size
        self.num_heads = num_heads
        self.proj = nn.Linear(hidden_size, hidden_size)

    def forward(self, x, *args, **kwargs):
        return self.proj(x), None

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes::

import torch.nn as nn
import torch.nn.functional as F


class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call :meth:to, etc.

Note

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool

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

Ancestors

torch.nn.modules.module.Module

Methods

def forward(self, x, *args, **kwargs) ‑> Callable[..., typing.Any]

Expand source code

def forward(self, x, *args, **kwargs):
    return self.proj(x), None

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the :class:Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class DummyBertModel

Expand source code

class DummyBertModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.config = SimpleNamespace(model_type="bert", hidden_size=64, num_attention_heads=4)
        layer = nn.Module()
        layer.attention = nn.Module()
        layer.attention.self = nn.Module()
        layer.attention.self.query = nn.Linear(64, 64)
        layer.attention.self.num_attention_heads = 4
        self.encoder = nn.Module()
        self.encoder.layer = nn.ModuleList([layer])

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes::

import torch.nn as nn
import torch.nn.functional as F


class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call :meth:to, etc.

Note

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool

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

Ancestors

torch.nn.modules.module.Module

class DummyDistilBertModel

Expand source code

class DummyDistilBertModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.config = SimpleNamespace(model_type="distilbert", hidden_size=64, num_attention_heads=4)
        layer = nn.Module()
        layer.attention = nn.Module()
        self.distilbert = nn.Module()
        self.distilbert.transformer = nn.Module()
        self.distilbert.transformer.layer = nn.ModuleList([layer])

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes::

import torch.nn as nn
import torch.nn.functional as F


class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call :meth:to, etc.

Note

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool

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

Ancestors

torch.nn.modules.module.Module

class DummyGPT2Model

Expand source code

class DummyGPT2Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.config = SimpleNamespace(model_type="gpt2", hidden_size=64, num_attention_heads=4)
        block = nn.Module()
        block.attn = nn.Module()
        block.attn.embed_dim = 64
        block.attn.num_heads = 4
        self.transformer = nn.Module()
        self.transformer.h = nn.ModuleList([block])

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes::

import torch.nn as nn
import torch.nn.functional as F


class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call :meth:to, etc.

Note

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool

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

Ancestors

torch.nn.modules.module.Module