【PPMix No.39】mPLUG-Owl3推理 (PaddlePaddle#865)

Co-authored-by: [email protected]@github.com <[email protected]>

paddlemix/examples/mPLUG_Owl3/README.md

@@ -0,0 +1,42 @@
+# mPLUG-Owl3
+
+## 1. 模型介绍
+
+**本仓库支持的模型权重:**
+
+| Model              |
+|--------------------|
+| mPLUG/mPLUG-Owl3-7B-241101  |
+
+注意：与huggingface权重同名，但权重为paddle框架的Tensor，使用`xxx.from_pretrained("mPLUG/mPLUG-Owl3-7B-241101")`即可自动下载该权重文件夹到缓存目录。
+
+
+## 2 环境准备
+
+1）[安装 PaddleMIX 环境依赖包](https://github.com/PaddlePaddle/PaddleMIX/tree/develop?tab=readme-ov-file#%E5%AE%89%E8%A3%85)
+
+2）pip install pillow tqdm paddlenlp==3.0.0b2
+
+注意：Python版本最好为3.10及以上版本。
+
+## 3 快速开始
+
+### 推理
+```bash
+# 图片理解
+CUDA_VISIBLE_DEVICES=0 python paddlemix/examples/mPLUG_Owl3/run_inference.py \
+```
+
+
+### 参考文献
+```BibTeX
+@misc{ye2024mplugowl3longimagesequenceunderstanding,
+      title={mPLUG-Owl3: Towards Long Image-Sequence Understanding in Multi-Modal Large Language Models},
+      author={Jiabo Ye and Haiyang Xu and Haowei Liu and Anwen Hu and Ming Yan and Qi Qian and Ji Zhang and Fei Huang and Jingren Zhou},
+      year={2024},
+      eprint={2408.04840},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV},
+      url={https://arxiv.org/abs/2408.04840},
+}
+```

paddlemix/examples/mPLUG_Owl3/requirement.txt

@@ -0,0 +1,3 @@
+pillow
+tqdm
+paddlenlp==3.0.0b2

paddlemix/examples/mPLUG_Owl3/run_inference.py

@@ -0,0 +1,46 @@
+# Copyright (c) 2024 PaddlePaddle Authors. 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.
+
+import paddle
+from paddlenlp.transformers import Qwen2Tokenizer
+from PIL import Image
+
+from paddlemix.models.mPLUGOwl3.configuration_mplugowl3 import mPLUGOwl3Config
+from paddlemix.models.mPLUGOwl3.modeling_mplugowl3 import mPLUGOwl3Model
+
+model_path = "mPLUG/mPLUG-Owl3-7B-241101"
+
+config = mPLUGOwl3Config.from_pretrained(model_path)
+model = mPLUGOwl3Model.from_pretrained(model_path, dtype=paddle.bfloat16).eval()
+tokenizer = Qwen2Tokenizer.from_pretrained(model_path)
+processor = model.init_processor(tokenizer)
+
+# image = Image.new('RGB', (500, 500), color='red')
+image = Image.open("paddlemix/demo_images/examples_image1.jpg").convert("RGB")
+
+messages = [{"role": "user", "content": """<|image|>Describe this image."""}, {"role": "assistant", "content": ""}]
+
+inputs = processor(messages, images=[image], videos=None)
+inputs["pixel_values"] = inputs["pixel_values"].cast(paddle.bfloat16)
+
+inputs.update(
+    {
+        "tokenizer": tokenizer,
+        "max_new_tokens": 512,  #
+        "decode_text": True,
+    }
+)
+
+res = model.generate(**inputs)
+print("output:\n", res)

paddlemix/models/mPLUGOwl3/__init__.py

@@ -0,0 +1,19 @@
+# Copyright (c) 2023 PaddlePaddle Authors. 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.
+
+from .configuration_hyper_qwen2 import *
+from .configuration_mplugowl3 import *
+from .modeling_hyper_qwen2 import *
+from .modeling_mplugowl3 import *
+from .modeling_navit_siglip import *

paddlemix/models/mPLUGOwl3/configuration_hyper_qwen2.py

@@ -0,0 +1,141 @@
+# Copyright (c) 2024 PaddlePaddle Authors. 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.
+
+from paddlenlp.transformers import PretrainedConfig
+
+
+class HyperQwen2Config(PretrainedConfig):
+    """
+    This is the configuration class to store the configuration of a [`Qwen2Model`]. It is used to instantiate a
+    Qwen2 model according to the specified arguments, defining the model architecture. Instantiating a configuration
+    with the defaults will yield a similar configuration to that of
+    Qwen2-7B-beta [Qwen/Qwen2-7B-beta](https://huggingface.co/Qwen/Qwen2-7B-beta).
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 151936):
+            Vocabulary size of the Qwen2 model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`Qwen2Model`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 22016):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*, defaults to 32):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `32`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 32768):
+            The maximum sequence length that this model might ever be used with.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-06):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether the model's input and output word embeddings should be tied.
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        use_sliding_window (`bool`, *optional*, defaults to `False`):
+            Whether to use sliding window attention.
+        sliding_window (`int`, *optional*, defaults to 4096):
+            Sliding window attention (SWA) window size. If not specified, will default to `4096`.
+        max_window_layers (`int`, *optional*, defaults to 28):
+            The number of layers that use SWA (Sliding Window Attention). The bottom layers use SWA while the top use full attention.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+
+    ```python
+    >>> from transformers import Qwen2Model, Qwen2Config
+
+    >>> # Initializing a Qwen2 style configuration
+    >>> configuration = Qwen2Config()
+
+    >>> # Initializing a model from the Qwen2-7B style configuration
+    >>> model = Qwen2Model(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "qwen2"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=151936,
+        hidden_size=4096,
+        intermediate_size=22016,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=32,
+        hidden_act="silu",
+        max_position_embeddings=32768,
+        initializer_range=0.02,
+        rms_norm_eps=1e-6,
+        use_cache=True,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        use_sliding_window=False,
+        sliding_window=4096,
+        max_window_layers=28,
+        attention_dropout=0.0,
+        hyper_layers=[1,9,17,25],
+        vision_batch_size=16,
+        rope_scaling=None,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.use_sliding_window = use_sliding_window
+        self.sliding_window = sliding_window if use_sliding_window else None
+        self.max_window_layers = max_window_layers
+        self.rope_scaling = rope_scaling
+        if self.rope_scaling is not None and "type" in self.rope_scaling:
+            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+        self.hyper_layers = hyper_layers
+        self.vision_batch_size = vision_batch_size
+        self.seq_length = 1 #self.max_length ###
+        super().__init__(
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )

paddlemix/models/mPLUGOwl3/configuration_mplugowl3.py

@@ -0,0 +1,54 @@
+# Copyright (c) 2024 PaddlePaddle Authors. 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.
+
+from paddlemix.utils.log import logger
+
+from .configuration_hyper_qwen2 import HyperQwen2Config
+from .modeling_navit_siglip import SigLipVisionConfig
+
+
+class mPLUGOwl3Config(HyperQwen2Config):
+    model_type = "mplugowl3"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    default_vision_config = {
+        "hidden_size": 1152,
+        "image_size": 378,
+        "intermediate_size": 4304,
+        "model_type": "siglip_vision_model",
+        "num_attention_heads": 16,
+        "num_hidden_layers": 27,
+        "patch_size": 14,
+    }
+
+    def __init__(
+        self,
+        use_cache=True,
+        vision_config=None,
+        **kwargs,
+    ):
+        self.use_cache = use_cache
+
+        # same as HuggingFaceM4/siglip-so400m-14-980-flash-attn2-navit add tgt_sizes
+        if vision_config is None:
+            self.vision_config = SigLipVisionConfig(**self.default_vision_config)
+            logger.info("vision_config is None, using default vision config")
+        elif isinstance(vision_config, dict):
+            self.vision_config = SigLipVisionConfig(**vision_config)
+        elif isinstance(vision_config, SigLipVisionConfig):
+            self.vision_config = vision_config
+        self.image_size = 378
+        self.patch_size = self.vision_config.patch_size
+
+        super().__init__(**kwargs)