add ConvNext & ConvNextv2 (#1201)

Author: zhtmike

mindone/transformers/__init__.py

@@ -255,6 +255,20 @@
     ConvBertLayer,
     ConvBertModel,
 )
+from .models.convnext import (
+    ConvNextBackbone,
+    ConvNextFeatureExtractor,
+    ConvNextForImageClassification,
+    ConvNextImageProcessor,
+    ConvNextModel,
+    ConvNextPreTrainedModel,
+)
+from .models.convnextv2 import (
+    ConvNextV2Backbone,
+    ConvNextV2ForImageClassification,
+    ConvNextV2Model,
+    ConvNextV2PreTrainedModel,
+)
 from .models.deberta import (
     DebertaForMaskedLM,
     DebertaForQuestionAnswering,

mindone/transformers/models/__init__.py

@@ -34,6 +34,8 @@
     clap,
     clip,
     convbert,
+    convnext,
+    convnextv2,
     depth_anything,
     dinov2,
     dpt,

mindone/transformers/models/auto/configuration_auto.py

@@ -46,6 +46,8 @@
         ("chameleon", "ChameleonConfig"),
         ("camembert", "CamembertConfig"),
         ("convbert", "ConvBertConfig"),
+        ("convnext", "ConvNextConfig"),
+        ("convnextv2", "ConvNextV2Config"),
         ("clip", "CLIPConfig"),
         ("clip_vision_model", "CLIPVisionConfig"),
         ("deberta", "DebertaConfig"),
@@ -143,6 +145,8 @@
         ("clip", "CLIP"),
         ("starcoder2", "Starcoder2"),
         ("clip_vision_model", "CLIPVisionModel"),
+        ("convnext", "ConvNeXT"),
+        ("convnextv2", "ConvNeXTV2"),
         ("deberta", "DeBERTa"),
         ("deberta-v2", "DeBERTa-v2"),
         ("depth_anything", "Depth Anything"),

mindone/transformers/models/auto/feature_extraction_auto.py

@@ -41,7 +41,11 @@
 
 logger = logging.get_logger(__name__)
 
-FEATURE_EXTRACTOR_MAPPING_NAMES = OrderedDict()
+FEATURE_EXTRACTOR_MAPPING_NAMES = OrderedDict(
+    [
+        ("convnext", "ConvNextFeatureExtractor"),
+    ]
+)
 
 FEATURE_EXTRACTOR_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, FEATURE_EXTRACTOR_MAPPING_NAMES)
 

mindone/transformers/models/auto/image_processing_auto.py

@@ -52,6 +52,8 @@
             ("blip-2", ("BlipImageProcessor",)),
             ("chameleon", ("ChameleonImageProcessor",)),
             ("clip", ("CLIPImageProcessor",)),
+            ("convnext", ("ConvNextImageProcessor",)),
+            ("convnextv2", ("ConvNextImageProcessor",)),
             ("depth_anything", ("DPTImageProcessor",)),
             ("dinov2", ("BitImageProcessor",)),
             ("dpt", ("DPTImageProcessor",)),

mindone/transformers/models/auto/modeling_auto.py

@@ -50,6 +50,8 @@
         ("clip", "CLIPModel"),
         ("clip_text_model", "CLIPTextModel"),
         ("clip_vision_model", "CLIPVisionModel"),
+        ("convnext", "ConvNextModel"),
+        ("convnextv2", "ConvNextV2Model"),
         ("deberta", "DebertaModel"),
         ("opt", "OPTModel"),
         ("deberta-v2", "DebertaV2Model"),
@@ -219,6 +221,8 @@
     [
         # Model for Image mapping
         ("bit", "BitModel"),
+        ("convnext", "ConvNextModel"),
+        ("convnextv2", "ConvNextV2Model"),
         ("detr", "DetrModel"),
         ("dinov2", "Dinov2Model"),
         ("dpt", "DPTModel"),
@@ -252,6 +256,8 @@
         # Model for Image Classification mapping
         ("bit", "BitForImageClassification"),
         ("clip", "CLIPForImageClassification"),
+        ("convnext", "ConvNextForImageClassification"),
+        ("convnextv2", "ConvNextV2ForImageClassification"),
         ("dinov2", "Dinov2ForImageClassification"),
         ("hiera", "HieraForImageClassification"),
         ("ijepa", "IJepaForImageClassification"),
@@ -601,6 +607,8 @@
 
 MODEL_FOR_BACKBONE_MAPPING_NAMES = OrderedDict(
     [
+        ("convnext", "ConvNextBackbone"),
+        ("convnextv2", "ConvNextV2Backbone"),
         ("dinov2", "Dinov2Backbone"),
         ("hiera", "HieraBackbone"),
         ("swin", "SwinBackbone"),

mindone/transformers/models/convnext/__init__.py

@@ -0,0 +1,19 @@
+# Copyright 2024 The HuggingFace Team. All rights reserved.
+#
+# This code is adapted from https://github.com/huggingface/transformers
+# with modifications to run transformers on mindspore.
+#
+# 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.
+from .feature_extraction_convnext import *
+from .image_processing_convnext import *
+from .modeling_convnext import *

mindone/transformers/models/convnext/feature_extraction_convnext.py

@@ -0,0 +1,38 @@
+# coding=utf-8
+# Copyright 2022 The HuggingFace Inc. team. All rights reserved.
+#
+# This code is adapted from https://github.com/huggingface/transformers
+# with modifications to run transformers on mindspore.
+#
+# 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.
+"""Feature extractor class for ConvNeXT."""
+
+import warnings
+
+from ...utils import logging
+from .image_processing_convnext import ConvNextImageProcessor
+
+logger = logging.get_logger(__name__)
+
+
+class ConvNextFeatureExtractor(ConvNextImageProcessor):
+    def __init__(self, *args, **kwargs) -> None:
+        warnings.warn(
+            "The class ConvNextFeatureExtractor is deprecated and will be removed in version 5 of Transformers."
+            " Please use ConvNextImageProcessor instead.",
+            FutureWarning,
+        )
+        super().__init__(*args, **kwargs)
+
+
+__all__ = ["ConvNextFeatureExtractor"]