Merge branch 'main' into bnb

docs/source/en/_toctree.yml

@@ -705,6 +705,8 @@
         title: Swin2SR
       - local: model_doc/table-transformer
         title: Table Transformer
+      - local: model_doc/timm_wrapper
+        title: Timm Wrapper
       - local: model_doc/upernet
         title: UperNet
       - local: model_doc/van

docs/source/en/index.md

@@ -321,6 +321,7 @@ Flax), PyTorch, and/or TensorFlow.
 |                         [TAPEX](model_doc/tapex)                         |       ✅        |         ✅         |      ✅      |
 |       [Time Series Transformer](model_doc/time_series_transformer)       |       ✅        |         ❌         |      ❌      |
 |                   [TimeSformer](model_doc/timesformer)                   |       ✅        |         ❌         |      ❌      |
+|                [TimmWrapperModel](model_doc/timm_wrapper)                |       ✅        |         ❌         |      ❌      |
 |        [Trajectory Transformer](model_doc/trajectory_transformer)        |       ✅        |         ❌         |      ❌      |
 |                  [Transformer-XL](model_doc/transfo-xl)                  |       ✅        |         ✅         |      ❌      |
 |                         [TrOCR](model_doc/trocr)                         |       ✅        |         ❌         |      ❌      |

docs/source/en/model_doc/timm_wrapper.md

@@ -0,0 +1,67 @@
+<!--Copyright 2024 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+# TimmWrapper
+
+## Overview
+
+Helper class to enable loading timm models to be used with the transformers library and its autoclasses.
+
+```python
+>>> import torch
+>>> from PIL import Image
+>>> from urllib.request import urlopen
+>>> from transformers import AutoModelForImageClassification, AutoImageProcessor
+
+>>> # Load image
+>>> image = Image.open(urlopen(
+...     'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
+... ))
+
+>>> # Load model and image processor
+>>> checkpoint = "timm/resnet50.a1_in1k"
+>>> image_processor = AutoImageProcessor.from_pretrained(checkpoint)
+>>> model = AutoModelForImageClassification.from_pretrained(checkpoint).eval()
+
+>>> # Preprocess image
+>>> inputs = image_processor(image)
+
+>>> # Forward pass
+>>> with torch.no_grad():
+...     logits = model(**inputs).logits
+
+>>> # Get top 5 predictions
+>>> top5_probabilities, top5_class_indices = torch.topk(logits.softmax(dim=1) * 100, k=5)
+```
+
+## TimmWrapperConfig
+
+[[autodoc]] TimmWrapperConfig
+
+## TimmWrapperImageProcessor
+
+[[autodoc]] TimmWrapperImageProcessor
+    - preprocess
+
+## TimmWrapperModel
+
+[[autodoc]] TimmWrapperModel
+    - forward
+
+## TimmWrapperForImageClassification
+
+[[autodoc]] TimmWrapperForImageClassification
+    - forward

examples/modular-transformers/modeling_dummy.py

@@ -228,9 +228,6 @@ def __init__(self, config: DummyConfig, layer_idx: Optional[int] = None):
         self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias)
         self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=config.attention_bias)
 
-        # TODO (joao): remove in v4.46 (RoPE is computed in the model, not in the decoder layers)
-        self.rotary_emb = DummyRotaryEmbedding(config=self.config)
-
     def forward(
         self,
         hidden_states: torch.Tensor,
@@ -240,7 +237,7 @@ def forward(
         output_attentions: bool = False,
         use_cache: bool = False,
         cache_position: Optional[torch.LongTensor] = None,
-        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
         **kwargs,
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
         bsz, q_len, _ = hidden_states.size()
@@ -254,16 +251,7 @@ def forward(
         key_states = key_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
         value_states = value_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
 
-        if position_embeddings is None:
-            logger.warning_once(
-                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
-                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
-                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
-                "removed and `position_embeddings` will be mandatory."
-            )
-            cos, sin = self.rotary_emb(value_states, position_ids)
-        else:
-            cos, sin = position_embeddings
+        cos, sin = position_embeddings
         query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
 
         if past_key_value is not None:
@@ -326,7 +314,7 @@ def forward(
         output_attentions: bool = False,
         use_cache: bool = False,
         cache_position: Optional[torch.LongTensor] = None,
-        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
         **kwargs: Unpack[FlashAttentionKwargs],
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
         if isinstance(past_key_value, StaticCache):
@@ -350,16 +338,7 @@ def forward(
         key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
         value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
 
-        if position_embeddings is None:
-            logger.warning_once(
-                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
-                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
-                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
-                "removed and `position_embeddings` will be mandatory."
-            )
-            cos, sin = self.rotary_emb(value_states, position_ids)
-        else:
-            cos, sin = position_embeddings
+        cos, sin = position_embeddings
         query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
 
         if past_key_value is not None:
@@ -441,7 +420,7 @@ def forward(
         output_attentions: bool = False,
         use_cache: bool = False,
         cache_position: Optional[torch.LongTensor] = None,
-        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
         **kwargs,
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
         if output_attentions:
@@ -472,16 +451,7 @@ def forward(
         key_states = key_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
         value_states = value_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
 
-        if position_embeddings is None:
-            logger.warning_once(
-                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
-                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
-                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
-                "removed and `position_embeddings` will be mandatory."
-            )
-            cos, sin = self.rotary_emb(value_states, position_ids)
-        else:
-            cos, sin = position_embeddings
+        cos, sin = position_embeddings
         query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
 
         if past_key_value is not None:
@@ -551,7 +521,7 @@ def forward(
         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,  # will become mandatory in v4.46
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
         **kwargs,
     ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
         """

examples/modular-transformers/modeling_multimodal1.py

@@ -228,9 +228,6 @@ def __init__(self, config: Multimodal1TextConfig, layer_idx: Optional[int] = Non
         self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias)
         self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=config.attention_bias)
 
-        # TODO (joao): remove in v4.46 (RoPE is computed in the model, not in the decoder layers)
-        self.rotary_emb = Multimodal1TextRotaryEmbedding(config=self.config)
-
     def forward(
         self,
         hidden_states: torch.Tensor,
@@ -240,7 +237,7 @@ def forward(
         output_attentions: bool = False,
         use_cache: bool = False,
         cache_position: Optional[torch.LongTensor] = None,
-        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
         **kwargs,
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
         bsz, q_len, _ = hidden_states.size()
@@ -254,16 +251,7 @@ def forward(
         key_states = key_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
         value_states = value_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
 
-        if position_embeddings is None:
-            logger.warning_once(
-                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
-                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
-                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
-                "removed and `position_embeddings` will be mandatory."
-            )
-            cos, sin = self.rotary_emb(value_states, position_ids)
-        else:
-            cos, sin = position_embeddings
+        cos, sin = position_embeddings
         query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
 
         if past_key_value is not None:
@@ -326,7 +314,7 @@ def forward(
         output_attentions: bool = False,
         use_cache: bool = False,
         cache_position: Optional[torch.LongTensor] = None,
-        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
         **kwargs: Unpack[FlashAttentionKwargs],
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
         if isinstance(past_key_value, StaticCache):
@@ -350,16 +338,7 @@ def forward(
         key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
         value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
 
-        if position_embeddings is None:
-            logger.warning_once(
-                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
-                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
-                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
-                "removed and `position_embeddings` will be mandatory."
-            )
-            cos, sin = self.rotary_emb(value_states, position_ids)
-        else:
-            cos, sin = position_embeddings
+        cos, sin = position_embeddings
         query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
 
         if past_key_value is not None:
@@ -441,7 +420,7 @@ def forward(
         output_attentions: bool = False,
         use_cache: bool = False,
         cache_position: Optional[torch.LongTensor] = None,
-        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
         **kwargs,
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
         if output_attentions:
@@ -472,16 +451,7 @@ def forward(
         key_states = key_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
         value_states = value_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
 
-        if position_embeddings is None:
-            logger.warning_once(
-                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
-                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
-                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
-                "removed and `position_embeddings` will be mandatory."
-            )
-            cos, sin = self.rotary_emb(value_states, position_ids)
-        else:
-            cos, sin = position_embeddings
+        cos, sin = position_embeddings
         query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
 
         if past_key_value is not None:
@@ -553,7 +523,7 @@ def forward(
         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,  # will become mandatory in v4.46
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
         **kwargs,
     ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
         """