SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

No more Tuple, List, Dict (#38797)

Browse Source 
* No more Tuple, List, Dict

* make fixup

* More style fixes

* Docstring fixes with regex replacement

* Trigger tests

* Redo fixes after rebase

* Fix copies

* [test all]

* update

* [test all]

* update

* [test all]

* make style after rebase

* Patch the hf_argparser test

* Patch the hf_argparser test

* style fixes

* style fixes

* style fixes

* Fix docstrings in Cohere test

* [test all]

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
This commit is contained in:
Matt
2025-06-17 19:37:18 +01:00
committed by GitHub
parent a396f4324b
commit 508a704055
1291 changed files with 14906 additions and 14941 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
examples/modular-transformers/configuration_my_new_model.py
View File

@@ -91,11 +91,11 @@ class MyNewModelConfig(PretrainedConfig):
`beta_slow` (`float`, *optional*):
Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear
ramp function. If unspecified, it defaults to 1.
`short_factor` (`List[float]`, *optional*):
`short_factor` (`list[float]`, *optional*):
Only used with 'longrope'. The scaling factor to be applied to short contexts (<
`original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
size divided by the number of attention heads divided by 2
`long_factor` (`List[float]`, *optional*):
`long_factor` (`list[float]`, *optional*):
Only used with 'longrope'. The scaling factor to be applied to long contexts (<
`original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
size divided by the number of attention heads divided by 2

28
examples/modular-transformers/image_processing_new_imgproc_model.py
View File

@@ -4,7 +4,7 @@
# the file from the modular. If any change should be done, please apply the change to the
# modular_new_imgproc_model.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from typing import Dict, List, Optional, Union
from typing import Optional, Union
import numpy as np
import torch
@@ -57,11 +57,11 @@ class ImgprocModelImageProcessor(BaseImageProcessor):
do_normalize (`bool`, *optional*, defaults to `True`):
Whether to normalize the image. Can be overridden by the `do_normalize` parameter in the `preprocess`
method. Can be overridden by the `do_normalize` parameter in the `preprocess` method.
image_mean (`float` or `List[float]`, *optional*, defaults to `IMAGENET_STANDARD_MEAN`):
image_mean (`float` or `list[float]`, *optional*, defaults to `IMAGENET_STANDARD_MEAN`):
Mean to use if normalizing the image. This is a float or list of floats the length of the number of
channels in the image. Can be overridden by the `image_mean` parameter in the `preprocess` method. Can be
overridden by the `image_mean` parameter in the `preprocess` method.
image_std (`float` or `List[float]`, *optional*, defaults to `IMAGENET_STANDARD_STD`):
image_std (`float` or `list[float]`, *optional*, defaults to `IMAGENET_STANDARD_STD`):
Standard deviation to use if normalizing the image. This is a float or list of floats the length of the
number of channels in the image. Can be overridden by the `image_std` parameter in the `preprocess` method.
Can be overridden by the `image_std` parameter in the `preprocess` method.
@@ -74,13 +74,13 @@ class ImgprocModelImageProcessor(BaseImageProcessor):
def __init__(
self,
do_resize: bool = True,
size: Optional[Dict[str, int]] = None,
size: Optional[dict[str, int]] = None,
resample: PILImageResampling = PILImageResampling.BICUBIC,
do_rescale: bool = True,
rescale_factor: Union[int, float] = 1 / 255,
do_normalize: bool = True,
image_mean: Optional[Union[float, List[float]]] = None,
image_std: Optional[Union[float, List[float]]] = None,
image_mean: Optional[Union[float, list[float]]] = None,
image_std: Optional[Union[float, list[float]]] = None,
do_convert_rgb: bool = True,
**kwargs,
) -> None:
@@ -101,7 +101,7 @@ class ImgprocModelImageProcessor(BaseImageProcessor):
def resize(
self,
image: np.ndarray,
size: Dict[str, int],
size: dict[str, int],
resample: PILImageResampling = PILImageResampling.BICUBIC,
data_format: Optional[Union[str, ChannelDimension]] = None,
input_data_format: Optional[Union[str, ChannelDimension]] = None,
@@ -113,7 +113,7 @@ class ImgprocModelImageProcessor(BaseImageProcessor):
Args:
image (`np.ndarray`):
Image to resize.
size (`Dict[str, int]`):
size (`dict[str, int]`):
Dictionary in the format `{"height": int, "width": int}` specifying the size of the output image.
resample (`PILImageResampling`, *optional*, defaults to `PILImageResampling.BICUBIC`):
`PILImageResampling` filter to use when resizing the image e.g. `PILImageResampling.BICUBIC`.
@@ -151,13 +151,13 @@ class ImgprocModelImageProcessor(BaseImageProcessor):
self,
images: ImageInput,
do_resize: Optional[bool] = None,
size: Optional[Dict[str, int]] = None,
size: Optional[dict[str, int]] = None,
resample: PILImageResampling = None,
do_rescale: Optional[bool] = None,
rescale_factor: Optional[float] = None,
do_normalize: Optional[bool] = None,
image_mean: Optional[Union[float, List[float]]] = None,
image_std: Optional[Union[float, List[float]]] = None,
image_mean: Optional[Union[float, list[float]]] = None,
image_std: Optional[Union[float, list[float]]] = None,
return_tensors: Optional[Union[str, TensorType]] = None,
do_convert_rgb: Optional[bool] = None,
data_format: ChannelDimension = ChannelDimension.FIRST,
@@ -172,7 +172,7 @@ class ImgprocModelImageProcessor(BaseImageProcessor):
passing in images with pixel values between 0 and 1, set `do_rescale=False`.
do_resize (`bool`, *optional*, defaults to `self.do_resize`):
Whether to resize the image.
size (`Dict[str, int]`, *optional*, defaults to `self.size`):
size (`dict[str, int]`, *optional*, defaults to `self.size`):
Controls the size of the image after `resize`. The shortest edge of the image is resized to
`size["shortest_edge"]` whilst preserving the aspect ratio. If the longest edge of this resized image
is > `int(size["shortest_edge"] * (1333 / 800))`, then the image is resized again to make the longest
@@ -185,9 +185,9 @@ class ImgprocModelImageProcessor(BaseImageProcessor):
Rescale factor to rescale the image by if `do_rescale` is set to `True`.
do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
Whether to normalize the image.
image_mean (`float` or `List[float]`, *optional*, defaults to `self.image_mean`):
image_mean (`float` or `list[float]`, *optional*, defaults to `self.image_mean`):
Image mean to normalize the image by if `do_normalize` is set to `True`.
image_std (`float` or `List[float]`, *optional*, defaults to `self.image_std`):
image_std (`float` or `list[float]`, *optional*, defaults to `self.image_std`):
Image standard deviation to normalize the image by if `do_normalize` is set to `True`.
do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`):
Whether to convert the image to RGB.

4
examples/modular-transformers/modeling_add_function.py
View File

@@ -5,7 +5,7 @@
# modular_add_function.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# Note that zamba does not have the `apply_rotary_pos_emb` function!
from typing import Optional, Tuple
from typing import Optional
import torch
from torch import nn
@@ -62,5 +62,5 @@ class TestAttention(nn.Module):
def __init__(self):
pass
def forward(self) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
def forward(self) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
_ = apply_rotary_pos_emb(1, 1, 1, 1)

10
examples/modular-transformers/modeling_dummy.py
View File

@@ -4,7 +4,7 @@
# the file from the modular. If any change should be done, please apply the change to the
# modular_dummy.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from typing import Callable, Optional, Tuple, Union
from typing import Callable, Optional, Union
import torch
from torch import nn
@@ -210,12 +210,12 @@ class DummyAttention(nn.Module):
def forward(
self,
hidden_states: torch.Tensor,
position_embeddings: Tuple[torch.Tensor, torch.Tensor],
position_embeddings: tuple[torch.Tensor, torch.Tensor],
attention_mask: Optional[torch.Tensor],
past_key_value: Optional[Cache] = None,
cache_position: Optional[torch.LongTensor] = None,
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
input_shape = hidden_states.shape[:-1]
hidden_shape = (*input_shape, -1, self.head_dim)
@@ -278,9 +278,9 @@ class DummyDecoderLayer(GradientCheckpointingLayer):
output_attentions: Optional[bool] = False,
use_cache: Optional[bool] = False,
cache_position: Optional[torch.LongTensor] = None,
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
) -> tuple[torch.FloatTensor, Optional[tuple[torch.FloatTensor, torch.FloatTensor]]]:
residual = hidden_states
hidden_states = self.input_layernorm(hidden_states)

22
examples/modular-transformers/modeling_dummy_bert.py
View File

@@ -6,7 +6,7 @@
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
import math
import os
from typing import Optional, Tuple, Union
from typing import Optional, Union
import torch
from packaging import version
@@ -136,9 +136,9 @@ class DummyBertSelfAttention(nn.Module):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_value: Optional[tuple[tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
) -> tuple[torch.Tensor]:
mixed_query_layer = self.query(hidden_states)
# If this is instantiated as a cross-attention module, the keys
@@ -245,9 +245,9 @@ class DummyBertSdpaSelfAttention(DummyBertSelfAttention):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_value: Optional[tuple[tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
) -> tuple[torch.Tensor]:
if self.position_embedding_type != "absolute" or output_attentions or head_mask is not None:
# TODO: Improve this warning with e.g. `model.config._attn_implementation = "manual"` once implemented.
logger.warning_once(
@@ -386,9 +386,9 @@ class DummyBertAttention(nn.Module):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_value: Optional[tuple[tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
) -> tuple[torch.Tensor]:
self_outputs = self.self(
hidden_states,
attention_mask,
@@ -454,9 +454,9 @@ class DummyBertLayer(nn.Module):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_value: Optional[tuple[tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
) -> tuple[torch.Tensor]:
# decoder uni-directional self-attention cached key/values tuple is at positions 1,2
self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
self_attention_outputs = self.attention(
@@ -532,12 +532,12 @@ class DummyBertEncoder(nn.Module):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_values: Optional[tuple[tuple[torch.FloatTensor]]] = None,
use_cache: Optional[bool] = None,
output_attentions: Optional[bool] = False,
output_hidden_states: Optional[bool] = False,
return_dict: Optional[bool] = True,
) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]:
) -> Union[tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]:
all_hidden_states = () if output_hidden_states else None
all_self_attentions = () if output_attentions else None
all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None

6
examples/modular-transformers/modeling_from_uppercase_model.py
View File

@@ -4,7 +4,7 @@
# the file from the modular. If any change should be done, please apply the change to the
# modular_from_uppercase_model.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from typing import Callable, Optional, Tuple, Union
from typing import Callable, Optional, Union
import torch
from torch import nn
@@ -71,7 +71,7 @@ class FromUppercaseModelAttention(nn.Module):
attention_mask: Optional[torch.Tensor] = None,
causal_attention_mask: Optional[torch.Tensor] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor]]:
"""Input shape: Batch x Time x Channel"""
batch_size, seq_length, embed_dim = hidden_states.shape
@@ -153,7 +153,7 @@ class FromUppercaseModelEncoderLayer(nn.Module):
attention_mask: torch.Tensor,
causal_attention_mask: torch.Tensor,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.FloatTensor]:
) -> tuple[torch.FloatTensor]:
"""
Args:
hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`

10
examples/modular-transformers/modeling_multimodal1.py
View File

@@ -4,7 +4,7 @@
# the file from the modular. If any change should be done, please apply the change to the
# modular_multimodal1.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from typing import Callable, Optional, Tuple, Union
from typing import Callable, Optional, Union
import torch
from torch import nn
@@ -210,12 +210,12 @@ class Multimodal1TextAttention(nn.Module):
def forward(
self,
hidden_states: torch.Tensor,
position_embeddings: Tuple[torch.Tensor, torch.Tensor],
position_embeddings: tuple[torch.Tensor, torch.Tensor],
attention_mask: Optional[torch.Tensor],
past_key_value: Optional[Cache] = None,
cache_position: Optional[torch.LongTensor] = None,
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
input_shape = hidden_states.shape[:-1]
hidden_shape = (*input_shape, -1, self.head_dim)
@@ -278,9 +278,9 @@ class Multimodal1TextDecoderLayer(GradientCheckpointingLayer):
output_attentions: Optional[bool] = False,
use_cache: Optional[bool] = False,
cache_position: Optional[torch.LongTensor] = None,
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
) -> tuple[torch.FloatTensor, Optional[tuple[torch.FloatTensor, torch.FloatTensor]]]:
residual = hidden_states
hidden_states = self.input_layernorm(hidden_states)

8
examples/modular-transformers/modeling_multimodal2.py
View File

@@ -5,7 +5,7 @@
# modular_multimodal2.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from typing import Callable, Optional, Tuple, Union
from typing import Callable, Optional, Union
import torch
from torch import nn
@@ -81,7 +81,7 @@ class Multimodal2VisionAttention(nn.Module):
attention_mask: Optional[torch.Tensor] = None,
causal_attention_mask: Optional[torch.Tensor] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor]]:
"""Input shape: Batch x Time x Channel"""
batch_size, seq_length, embed_dim = hidden_states.shape
@@ -177,7 +177,7 @@ class Multimodal2Attention(nn.Module):
attention_mask: Optional[torch.Tensor] = None,
causal_attention_mask: Optional[torch.Tensor] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor]]:
"""Input shape: Batch x Time x Channel"""
batch_size, seq_length, embed_dim = hidden_states.shape
@@ -244,7 +244,7 @@ class Multimodal2VisionEncoderLayer(nn.Module):
attention_mask: torch.Tensor,
causal_attention_mask: torch.Tensor,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.FloatTensor]:
) -> tuple[torch.FloatTensor]:
"""
Args:
hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`

12
examples/modular-transformers/modeling_my_new_model2.py
View File

@@ -4,7 +4,7 @@
# the file from the modular. If any change should be done, please apply the change to the
# modular_my_new_model2.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from typing import Callable, List, Optional, Tuple, Union
from typing import Callable, Optional, Union
import torch
from torch import nn
@@ -208,12 +208,12 @@ class MyNewModel2Attention(nn.Module):
def forward(
self,
hidden_states: torch.Tensor,
position_embeddings: Tuple[torch.Tensor, torch.Tensor],
position_embeddings: tuple[torch.Tensor, torch.Tensor],
attention_mask: Optional[torch.Tensor],
past_key_value: Optional[Cache] = None,
cache_position: Optional[torch.LongTensor] = None,
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
input_shape = hidden_states.shape[:-1]
hidden_shape = (*input_shape, -1, self.head_dim)
@@ -276,9 +276,9 @@ class MyNewModel2DecoderLayer(GradientCheckpointingLayer):
output_attentions: Optional[bool] = False,
use_cache: Optional[bool] = False,
cache_position: Optional[torch.LongTensor] = None,
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
) -> tuple[torch.FloatTensor, Optional[tuple[torch.FloatTensor, torch.FloatTensor]]]:
residual = hidden_states
hidden_states = self.input_layernorm(hidden_states)
@@ -469,7 +469,7 @@ class MyNewModel2Model(MyNewModel2PreTrainedModel):
input_ids: Optional[torch.LongTensor] = None,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None,
past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]] = None,
inputs_embeds: Optional[torch.FloatTensor] = None,
use_cache: Optional[bool] = None,
output_attentions: Optional[bool] = None,

14
examples/modular-transformers/modeling_new_task_model.py
View File

@@ -5,7 +5,7 @@
# modular_new_task_model.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from dataclasses import dataclass
from typing import ClassVar, List, Optional, Tuple, Union
from typing import ClassVar, Optional, Union
import torch
from torch import nn
@@ -88,9 +88,9 @@ class NewTaskModelCausalLMOutputWithPast(ModelOutput):
loss: Optional[torch.FloatTensor] = None
logits: Optional[torch.FloatTensor] = None
past_key_values: Optional[Union[List[torch.FloatTensor], Cache]] = None
hidden_states: Optional[Tuple[torch.FloatTensor]] = None
attentions: Optional[Tuple[torch.FloatTensor]] = None
past_key_values: Optional[Union[list[torch.FloatTensor], Cache]] = None
hidden_states: Optional[tuple[torch.FloatTensor]] = None
attentions: Optional[tuple[torch.FloatTensor]] = None
image_hidden_states: Optional[torch.FloatTensor] = None
@@ -249,7 +249,7 @@ class NewTaskModelModel(NewTaskModelPreTrainedModel):
pixel_values: torch.FloatTensor = None,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_values: Optional[Union[List[torch.FloatTensor], Cache]] = None,
past_key_values: Optional[Union[list[torch.FloatTensor], Cache]] = None,
token_type_ids: Optional[torch.LongTensor] = None,
cache_position: Optional[torch.LongTensor] = None,
inputs_embeds: Optional[torch.FloatTensor] = None,
@@ -259,7 +259,7 @@ class NewTaskModelModel(NewTaskModelPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
**kwargs: Unpack[FlashAttentionKwargs],
) -> Union[Tuple, NewTaskModelModelOutputWithPast]:
) -> Union[tuple, NewTaskModelModelOutputWithPast]:
r"""
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
@@ -442,7 +442,7 @@ class NewTaskModelForNewTask(NewTaskModelPreTrainedModel, GenerationMixin):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
num_logits_to_keep: int = 0,
) -> Union[Tuple, NewTaskModelCausalLMOutputWithPast]:
) -> Union[tuple, NewTaskModelCausalLMOutputWithPast]:
r"""
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,

26
examples/modular-transformers/modeling_roberta.py
View File

@@ -6,7 +6,7 @@
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
import math
import os
from typing import List, Optional, Tuple, Union
from typing import Optional, Union
import torch
import torch.nn as nn
@@ -139,9 +139,9 @@ class RobertaSelfAttention(nn.Module):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_value: Optional[tuple[tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
) -> tuple[torch.Tensor]:
mixed_query_layer = self.query(hidden_states)
# If this is instantiated as a cross-attention module, the keys
@@ -248,9 +248,9 @@ class RobertaSdpaSelfAttention(RobertaSelfAttention):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_value: Optional[tuple[tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
) -> tuple[torch.Tensor]:
if self.position_embedding_type != "absolute" or output_attentions or head_mask is not None:
# TODO: Improve this warning with e.g. `model.config._attn_implementation = "manual"` once implemented.
logger.warning_once(
@@ -389,9 +389,9 @@ class RobertaAttention(nn.Module):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_value: Optional[tuple[tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
) -> tuple[torch.Tensor]:
self_outputs = self.self(
hidden_states,
attention_mask,
@@ -457,9 +457,9 @@ class RobertaLayer(nn.Module):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_value: Optional[tuple[tuple[torch.FloatTensor]]] = None,
output_attentions: Optional[bool] = False,
) -> Tuple[torch.Tensor]:
) -> tuple[torch.Tensor]:
# decoder uni-directional self-attention cached key/values tuple is at positions 1,2
self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
self_attention_outputs = self.attention(
@@ -535,12 +535,12 @@ class RobertaEncoder(nn.Module):
head_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
past_key_values: Optional[tuple[tuple[torch.FloatTensor]]] = None,
use_cache: Optional[bool] = None,
output_attentions: Optional[bool] = False,
output_hidden_states: Optional[bool] = False,
return_dict: Optional[bool] = True,
) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]:
) -> Union[tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]:
all_hidden_states = () if output_hidden_states else None
all_self_attentions = () if output_attentions else None
all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
@@ -903,12 +903,12 @@ class RobertaModel(RobertaPreTrainedModel):
inputs_embeds: Optional[torch.Tensor] = None,
encoder_hidden_states: Optional[torch.Tensor] = None,
encoder_attention_mask: Optional[torch.Tensor] = None,
past_key_values: Optional[List[torch.FloatTensor]] = None,
past_key_values: Optional[list[torch.FloatTensor]] = None,
use_cache: Optional[bool] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
) -> Union[tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
r"""
encoder_hidden_states (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if

10
examples/modular-transformers/modeling_super.py
View File

@@ -4,7 +4,7 @@
# the file from the modular. If any change should be done, please apply the change to the
# modular_super.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from typing import Callable, Optional, Tuple, Union
from typing import Callable, Optional, Union
import torch
from torch import nn
@@ -211,12 +211,12 @@ class SuperAttention(nn.Module):
def forward(
self,
hidden_states: torch.Tensor,
position_embeddings: Tuple[torch.Tensor, torch.Tensor],
position_embeddings: tuple[torch.Tensor, torch.Tensor],
attention_mask: Optional[torch.Tensor],
past_key_value: Optional[Cache] = None,
cache_position: Optional[torch.LongTensor] = None,
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
input_shape = hidden_states.shape[:-1]
hidden_shape = (*input_shape, -1, self.head_dim)
@@ -279,9 +279,9 @@ class SuperDecoderLayer(GradientCheckpointingLayer):
output_attentions: Optional[bool] = False,
use_cache: Optional[bool] = False,
cache_position: Optional[torch.LongTensor] = None,
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
) -> tuple[torch.FloatTensor, Optional[tuple[torch.FloatTensor, torch.FloatTensor]]]:
residual = hidden_states
hidden_states = self.input_layernorm(hidden_states)

6
examples/modular-transformers/modeling_switch_function.py
View File

@@ -5,7 +5,7 @@
# modular_switch_function.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# Note that llama and cohere have different definitions for rotate_half
from typing import Callable, Optional, Tuple
from typing import Callable, Optional
import torch
from torch import nn
@@ -123,12 +123,12 @@ class SwitchFunctionAttention(nn.Module):
def forward(
self,
hidden_states: torch.Tensor,
position_embeddings: Tuple[torch.Tensor, torch.Tensor],
position_embeddings: tuple[torch.Tensor, torch.Tensor],
attention_mask: Optional[torch.Tensor],
past_key_value: Optional[Cache] = None,
cache_position: Optional[torch.LongTensor] = None,
**kwargs: Unpack[FlashAttentionKwargs],
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
input_shape = hidden_states.shape[:-1]
hidden_shape = (*input_shape, -1, self.head_dim)

26
examples/modular-transformers/modeling_test_detr.py
View File

@@ -7,7 +7,7 @@
import math
import warnings
from dataclasses import dataclass
from typing import List, Optional, Tuple, Union
from typing import Optional, Union
import torch
import torch.nn.functional as F
@@ -43,7 +43,7 @@ class MultiScaleDeformableAttention(nn.Module):
self,
value: Tensor,
value_spatial_shapes: Tensor,
value_spatial_shapes_list: List[Tuple],
value_spatial_shapes_list: list[tuple],
level_start_index: Tensor,
sampling_locations: Tensor,
attention_weights: Tensor,
@@ -124,9 +124,9 @@ class TestDetrDecoderOutput(ModelOutput):
last_hidden_state: Optional[torch.FloatTensor] = None
intermediate_hidden_states: Optional[torch.FloatTensor] = None
intermediate_reference_points: Optional[torch.FloatTensor] = None
hidden_states: Optional[Tuple[torch.FloatTensor]] = None
attentions: Optional[Tuple[torch.FloatTensor]] = None
cross_attentions: Optional[Tuple[torch.FloatTensor]] = None
hidden_states: Optional[tuple[torch.FloatTensor]] = None
attentions: Optional[tuple[torch.FloatTensor]] = None
cross_attentions: Optional[tuple[torch.FloatTensor]] = None
@dataclass
@@ -177,12 +177,12 @@ class TestDetrModelOutput(ModelOutput):
last_hidden_state: Optional[torch.FloatTensor] = None
intermediate_hidden_states: Optional[torch.FloatTensor] = None
intermediate_reference_points: Optional[torch.FloatTensor] = None
decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
decoder_attentions: Optional[Tuple[torch.FloatTensor]] = None
cross_attentions: Optional[Tuple[torch.FloatTensor]] = None
decoder_hidden_states: Optional[tuple[torch.FloatTensor]] = None
decoder_attentions: Optional[tuple[torch.FloatTensor]] = None
cross_attentions: Optional[tuple[torch.FloatTensor]] = None
encoder_last_hidden_state: Optional[torch.FloatTensor] = None
encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
encoder_attentions: Optional[Tuple[torch.FloatTensor]] = None
encoder_hidden_states: Optional[tuple[torch.FloatTensor]] = None
encoder_attentions: Optional[tuple[torch.FloatTensor]] = None
enc_outputs_class: Optional[torch.FloatTensor] = None
enc_outputs_coord_logits: Optional[torch.FloatTensor] = None
@@ -557,7 +557,7 @@ class TestDetrMultiheadAttention(nn.Module):
attention_mask: Optional[torch.Tensor] = None,
position_embeddings: Optional[torch.Tensor] = None,
output_attentions: bool = False,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
"""Input shape: Batch x Time x Channel"""
batch_size, target_len, embed_dim = hidden_states.size()
@@ -1431,7 +1431,7 @@ class TestDetrModel(TestDetrPreTrainedModel):
Args:
enc_output (Tensor[batch_size, sequence_length, hidden_size]): Output of the encoder.
padding_mask (Tensor[batch_size, sequence_length]): Padding mask for `enc_output`.
spatial_shapes (List[Tuple[int, int]]): Spatial shapes of the feature maps.
spatial_shapes (list[tuple[int, int]]): Spatial shapes of the feature maps.
Returns:
`tuple(torch.FloatTensor)`: A tuple of feature map and bbox prediction.
@@ -1499,7 +1499,7 @@ class TestDetrModel(TestDetrPreTrainedModel):
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
) -> Union[Tuple[torch.FloatTensor], TestDetrModelOutput]:
) -> Union[tuple[torch.FloatTensor], TestDetrModelOutput]:
r"""
Returns: