Chinese-CLIP

Chinese-CLIP

Overview

The Chinese-CLIP model was proposed in Chinese CLIP: Contrastive Vision-Language Pretraining in Chinese by An Yang, Junshu Pan, Junyang Lin, Rui Men, Yichang Zhang, Jingren Zhou, Chang Zhou. Chinese-CLIP is an implementation of CLIP (Radford et al., 2021) on a large-scale dataset of Chinese image-text pairs. It is capable of performing cross-modal retrieval and also playing as a vision backbone for vision tasks like zero-shot image classification, open-domain object detection, etc. The original Chinese-CLIP code is released at this link.

The abstract from the paper is the following:

The tremendous success of CLIP (Radford et al., 2021) has promoted the research and application of contrastive learning for vision-language pretraining. In this work, we construct a large-scale dataset of image-text pairs in Chinese, where most data are retrieved from publicly available datasets, and we pretrain Chinese CLIP models on the new dataset. We develop 5 Chinese CLIP models of multiple sizes, spanning from 77 to 958 million parameters. Furthermore, we propose a two-stage pretraining method, where the model is first trained with the image encoder frozen and then trained with all parameters being optimized, to achieve enhanced model performance. Our comprehensive experiments demonstrate that Chinese CLIP can achieve the state-of-the-art performance on MUGE, Flickr30K-CN, and COCO-CN in the setups of zero-shot learning and finetuning, and it is able to achieve competitive performance in zero-shot image classification based on the evaluation on the ELEVATER benchmark (Li et al., 2022). Our codes, pretrained models, and demos have been released.

Usage

The code snippet below shows how to compute image & text features and similarities:

Copied

>>> from PIL import Image
>>> import requests
>>> from transformers import ChineseCLIPProcessor, ChineseCLIPModel

>>> model = ChineseCLIPModel.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")
>>> processor = ChineseCLIPProcessor.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")

>>> url = "https://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/pokemon.jpeg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> # Squirtle, Bulbasaur, Charmander, Pikachu in English
>>> texts = ["ๆฐๅฐผ้พŸ", "ๅฆ™่›™็งๅญ", "ๅฐ็ซ้พ™", "็šฎๅกไธ˜"]

>>> # compute image feature
>>> inputs = processor(images=image, return_tensors="pt")
>>> image_features = model.get_image_features(**inputs)
>>> image_features = image_features / image_features.norm(p=2, dim=-1, keepdim=True)  # normalize

>>> # compute text features
>>> inputs = processor(text=texts, padding=True, return_tensors="pt")
>>> text_features = model.get_text_features(**inputs)
>>> text_features = text_features / text_features.norm(p=2, dim=-1, keepdim=True)  # normalize

>>> # compute image-text similarity scores
>>> inputs = processor(text=texts, images=image, return_tensors="pt", padding=True)
>>> outputs = model(**inputs)
>>> logits_per_image = outputs.logits_per_image  # this is the image-text similarity score
>>> probs = logits_per_image.softmax(dim=1)  # probs: [[1.2686e-03, 5.4499e-02, 6.7968e-04, 9.4355e-01]]

Currently, we release the following scales of pretrained Chinese-CLIP models at HF Model Hub:

The Chinese-CLIP model was contributed by OFA-Sys.

ChineseCLIPConfig

class transformers.ChineseCLIPConfig

<source>

( text_config = Nonevision_config = Noneprojection_dim = 512logit_scale_init_value = 2.6592**kwargs )

Parameters

  • text_config (dict, optional) โ€” Dictionary of configuration options used to initialize ChineseCLIPTextConfig.

  • vision_config (dict, optional) โ€” Dictionary of configuration options used to initialize ChineseCLIPVisionConfig.

  • projection_dim (int, optional, defaults to 512) โ€” Dimentionality of text and vision projection layers.

  • logit_scale_init_value (float, optional, defaults to 2.6592) โ€” The inital value of the logit_scale paramter. Default is used as per the original ChineseCLIP implementation.

  • kwargs (optional) โ€” Dictionary of keyword arguments.

ChineseCLIPConfig is the configuration class to store the configuration of a ChineseCLIPModel. It is used to instantiate Chinese-CLIP model according to the specified arguments, defining the text model and vision model configs. Instantiating a configuration with the defaults will yield a similar configuration to that of the Chinese-CLIP OFA-Sys/chinese-clip-vit-base-patch16 architecture.

Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.

Example:

Copied

>>> from transformers import ChineseCLIPConfig, ChineseCLIPModel

>>> # Initializing a ChineseCLIPConfig with OFA-Sys/chinese-clip-vit-base-patch16 style configuration
>>> configuration = ChineseCLIPConfig()

>>> # Initializing a ChineseCLIPModel (with random weights) from the OFA-Sys/chinese-clip-vit-base-patch16 style configuration
>>> model = ChineseCLIPModel(configuration)

>>> # Accessing the model configuration
>>> configuration = model.config

>>> # We can also initialize a ChineseCLIPConfig from a ChineseCLIPTextConfig and a ChineseCLIPVisionConfig

>>> # Initializing a ChineseCLIPTextConfig and ChineseCLIPVisionConfig configuration
>>> config_text = ChineseCLIPTextConfig()
>>> config_vision = ChineseCLIPVisionConfig()

>>> config = ChineseCLIPConfig.from_text_vision_configs(config_text, config_vision)

from_text_vision_configs

<source>

( text_config: ChineseCLIPTextConfigvision_config: ChineseCLIPVisionConfig**kwargs )

Instantiate a ChineseCLIPConfig (or a derived class) from Chinese-CLIP text model configuration and Chinese-CLIP vision model configuration. Returns: ChineseCLIPConfig: An instance of a configuration object

ChineseCLIPTextConfig

class transformers.ChineseCLIPTextConfig

<source>

( vocab_size = 30522hidden_size = 768num_hidden_layers = 12num_attention_heads = 12intermediate_size = 3072hidden_act = 'gelu'hidden_dropout_prob = 0.1attention_probs_dropout_prob = 0.1max_position_embeddings = 512type_vocab_size = 2initializer_range = 0.02initializer_factor = 1.0layer_norm_eps = 1e-12pad_token_id = 0position_embedding_type = 'absolute'use_cache = True**kwargs )

Parameters

  • vocab_size (int, optional, defaults to 30522) โ€” Vocabulary size of the CHINESE_CLIP model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling ChineseCLIPModel.

  • hidden_size (int, optional, defaults to 768) โ€” Dimensionality of the encoder layers and the pooler layer.

  • num_hidden_layers (int, optional, defaults to 12) โ€” Number of hidden layers in the Transformer encoder.

  • num_attention_heads (int, optional, defaults to 12) โ€” Number of attention heads for each attention layer in the Transformer encoder.

  • intermediate_size (int, optional, defaults to 3072) โ€” Dimensionality of the โ€œintermediateโ€ (often named feed-forward) layer in the Transformer encoder.

  • hidden_act (str or Callable, optional, defaults to "gelu") โ€” The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu", "relu", "silu" and "gelu_new" are supported.

  • hidden_dropout_prob (float, optional, defaults to 0.1) โ€” The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.

  • attention_probs_dropout_prob (float, optional, defaults to 0.1) โ€” The dropout ratio for the attention probabilities.

  • max_position_embeddings (int, optional, defaults to 512) โ€” The maximum sequence length that this model might ever be used with. Typically set this to something large just in case (e.g., 512 or 1024 or 2048).

  • type_vocab_size (int, optional, defaults to 2) โ€” The vocabulary size of the token_type_ids passed when calling ChineseCLIPModel.

  • initializer_range (float, optional, defaults to 0.02) โ€” The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

  • layer_norm_eps (float, optional, defaults to 1e-12) โ€” The epsilon used by the layer normalization layers.

  • position_embedding_type (str, optional, defaults to "absolute") โ€” Type of position embedding. Choose one of "absolute", "relative_key", "relative_key_query". For positional embeddings use "absolute". For more information on "relative_key", please refer to Self-Attention with Relative Position Representations (Shaw et al.). For more information on "relative_key_query", please refer to Method 4 in Improve Transformer Models with Better Relative Position Embeddings (Huang et al.).

  • use_cache (bool, optional, defaults to True) โ€” Whether or not the model should return the last key/values attentions (not used by all models). Only relevant if config.is_decoder=True.

This is the configuration class to store the configuration of a ChineseCLIPModel. It is used to instantiate a Chinese CLIP model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the Chinese CLIP [OFA-Sys/chinese-clip-vit-base-patch16](https: //boincai.com/OFA-Sys/chinese-clip-vit-base-patch16) architecture.

Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.

Example:

Copied

>>> from transformers import ChineseCLIPTextConfig, ChineseCLIPTextModel

>>> # Initializing a ChineseCLIPTextConfig with OFA-Sys/chinese-clip-vit-base-patch16 style configuration
>>> configuration = ChineseCLIPTextConfig()

>>> # Initializing a ChineseCLIPTextModel (with random weights) from the OFA-Sys/chinese-clip-vit-base-patch16 style configuration
>>> model = ChineseCLIPTextModel(configuration)

>>> # Accessing the model configuration
>>> configuration = model.config

ChineseCLIPVisionConfig

class transformers.ChineseCLIPVisionConfig

<source>

( hidden_size = 768intermediate_size = 3072projection_dim = 512num_hidden_layers = 12num_attention_heads = 12num_channels = 3image_size = 224patch_size = 32hidden_act = 'quick_gelu'layer_norm_eps = 1e-05attention_dropout = 0.0initializer_range = 0.02initializer_factor = 1.0**kwargs )

Parameters

  • hidden_size (int, optional, defaults to 768) โ€” Dimensionality of the encoder layers and the pooler layer.

  • intermediate_size (int, optional, defaults to 3072) โ€” Dimensionality of the โ€œintermediateโ€ (i.e., feed-forward) layer in the Transformer encoder.

  • num_hidden_layers (int, optional, defaults to 12) โ€” Number of hidden layers in the Transformer encoder.

  • num_attention_heads (int, optional, defaults to 12) โ€” Number of attention heads for each attention layer in the Transformer encoder.

  • image_size (int, optional, defaults to 224) โ€” The size (resolution) of each image.

  • patch_size (int, optional, defaults to 32) โ€” The size (resolution) of each patch.

  • hidden_act (str or function, optional, defaults to "quick_gelu") โ€” The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu", "relu", "selu" and "gelu_new" `"quick_gelu" are supported.

  • layer_norm_eps (float, optional, defaults to 1e-5) โ€” The epsilon used by the layer normalization layers.

  • attention_dropout (float, optional, defaults to 0.0) โ€” The dropout ratio for the attention probabilities.

  • initializer_range (float, optional, defaults to 0.02) โ€” The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

  • initializer_factor (`floatโ€œ, optional, defaults to 1) โ€” A factor for initializing all weight matrices (should be kept to 1, used internally for initialization testing).

This is the configuration class to store the configuration of a ChineseCLIPModel. It is used to instantiate an ChineseCLIP model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the ChineseCLIP [OFA-Sys/chinese-clip-vit-base-patch16](https: //boincai.com/OFA-Sys/chinese-clip-vit-base-patch16) architecture.

Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.

Example:

Copied

>>> from transformers import ChineseCLIPVisionConfig, ChineseCLIPVisionModel

>>> # Initializing a ChineseCLIPVisionConfig with OFA-Sys/chinese-clip-vit-base-patch16 style configuration
>>> configuration = ChineseCLIPVisionConfig()

>>> # Initializing a ChineseCLIPVisionModel (with random weights) from the OFA-Sys/chinese-clip-vit-base-patch16 style configuration
>>> model = ChineseCLIPVisionModel(configuration)

>>> # Accessing the model configuration
>>> configuration = model.config

ChineseCLIPImageProcessor

class transformers.ChineseCLIPImageProcessor

<source>

( do_resize: bool = Truesize: typing.Dict[str, int] = Noneresample: Resampling = <Resampling.BICUBIC: 3>do_center_crop: bool = Truecrop_size: typing.Dict[str, int] = Nonedo_rescale: bool = Truerescale_factor: typing.Union[int, float] = 0.00392156862745098do_normalize: bool = Trueimage_mean: typing.Union[float, typing.List[float], NoneType] = Noneimage_std: typing.Union[float, typing.List[float], NoneType] = Nonedo_convert_rgb: bool = True**kwargs )

Parameters

  • do_resize (bool, optional, defaults to True) โ€” Whether to resize the imageโ€™s (height, width) dimensions to the specified size. Can be overridden by do_resize in the preprocess method.

  • size (Dict[str, int] optional, defaults to {"shortest_edge" -- 224}): Size of the image after resizing. The shortest edge of the image is resized to size[โ€œshortest_edgeโ€], with the longest edge resized to keep the input aspect ratio. Can be overridden by size in the preprocess method.

  • resample (PILImageResampling, optional, defaults to PILImageResampling.BICUBIC) โ€” Resampling filter to use if resizing the image. Can be overridden by resample in the preprocess method.

  • do_center_crop (bool, optional, defaults to True) โ€” Whether to center crop the image to the specified crop_size. Can be overridden by do_center_crop in the preprocess method.

  • crop_size (Dict[str, int] optional, defaults to 224) โ€” Size of the output image after applying center_crop. Can be overridden by crop_size in the preprocess method.

  • do_rescale (bool, optional, defaults to True) โ€” Whether to rescale the image by the specified scale rescale_factor. Can be overridden by do_rescale in the preprocess method.

  • rescale_factor (int or float, optional, defaults to 1/255) โ€” Scale factor to use if rescaling the image. Can be overridden by rescale_factor in the preprocess method. do_normalize โ€” Whether to normalize the image. Can be overridden by do_normalize in the preprocess method.

  • image_mean (float or List[float], optional, defaults to IMAGENET_STANDARD_MEAN) โ€” 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.

  • image_std (float or List[float], optional, defaults to IMAGENET_STANDARD_STD) โ€” 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.

  • do_convert_rgb (bool, optional, defaults to True) โ€” Whether to convert the image to RGB.

Constructs a Chinese-CLIP image processor.

preprocess

<source>

( images: typing.Union[ForwardRef('PIL.Image.Image'), numpy.ndarray, ForwardRef('torch.Tensor'), typing.List[ForwardRef('PIL.Image.Image')], typing.List[numpy.ndarray], typing.List[ForwardRef('torch.Tensor')]]do_resize: bool = Nonesize: typing.Dict[str, int] = Noneresample: Resampling = Nonedo_center_crop: bool = Nonecrop_size: int = Nonedo_rescale: bool = Nonerescale_factor: float = Nonedo_normalize: bool = Noneimage_mean: typing.Union[float, typing.List[float], NoneType] = Noneimage_std: typing.Union[float, typing.List[float], NoneType] = Nonedo_convert_rgb: bool = Nonereturn_tensors: typing.Union[str, transformers.utils.generic.TensorType, NoneType] = Nonedata_format: typing.Optional[transformers.image_utils.ChannelDimension] = <ChannelDimension.FIRST: 'channels_first'>input_data_format: typing.Union[str, transformers.image_utils.ChannelDimension, NoneType] = None**kwargs )

Parameters

  • images (ImageInput) โ€” Image to preprocess. Expects a single or batch of images with pixel values ranging from 0 to 255. If passing in images with pixel values between 0 and 1, set do_rescale=False.

  • do_resize (bool, optional, defaults to self.do_resize) โ€” Whether to resize the image.

  • size (Dict[str, int], optional, defaults to self.size) โ€” Size of the image after resizing. Shortest edge of the image is resized to size[โ€œshortest_edgeโ€], with the longest edge resized to keep the input aspect ratio.

  • resample (int, optional, defaults to self.resample) โ€” Resampling filter to use if resizing the image. This can be one of the enum PILImageResampling. Only has an effect if do_resize is set to True.

  • do_center_crop (bool, optional, defaults to self.do_center_crop) โ€” Whether to center crop the image.

  • crop_size (Dict[str, int], optional, defaults to self.crop_size) โ€” Size of the center crop. Only has an effect if do_center_crop is set to True.

  • do_rescale (bool, optional, defaults to self.do_rescale) โ€” Whether to rescale the image.

  • rescale_factor (float, optional, defaults to self.rescale_factor) โ€” Rescale factor to rescale the image by if do_rescale is set to True.

  • do_normalize (bool, optional, defaults to self.do_normalize) โ€” Whether to normalize the image.

  • image_mean (float or List[float], optional, defaults to self.image_mean) โ€” Image mean to use for normalization. Only has an effect if do_normalize is set to True.

  • image_std (float or List[float], optional, defaults to self.image_std) โ€” Image standard deviation to use for normalization. Only has an effect if do_normalize is set to True.

  • do_convert_rgb (bool, optional, defaults to self.do_convert_rgb) โ€” Whether to convert the image to RGB.

  • return_tensors (str or TensorType, optional) โ€” The type of tensors to return. Can be one of:

    • Unset: Return a list of np.ndarray.

    • TensorType.TENSORFLOW or 'tf': Return a batch of type tf.Tensor.

    • TensorType.PYTORCH or 'pt': Return a batch of type torch.Tensor.

    • TensorType.NUMPY or 'np': Return a batch of type np.ndarray.

    • TensorType.JAX or 'jax': Return a batch of type jax.numpy.ndarray.

  • data_format (ChannelDimension or str, optional, defaults to ChannelDimension.FIRST) โ€” The channel dimension format for the output image. Can be one of:

    • "channels_first" or ChannelDimension.FIRST: image in (num_channels, height, width) format.

    • "channels_last" or ChannelDimension.LAST: image in (height, width, num_channels) format.

    • Unset: Use the channel dimension format of the input image.

  • input_data_format (ChannelDimension or str, optional) โ€” The channel dimension format for the input image. If unset, the channel dimension format is inferred from the input image. Can be one of:

    • "channels_first" or ChannelDimension.FIRST: image in (num_channels, height, width) format.

    • "channels_last" or ChannelDimension.LAST: image in (height, width, num_channels) format.

    • "none" or ChannelDimension.NONE: image in (height, width) format.

Preprocess an image or batch of images.

ChineseCLIPFeatureExtractor

class transformers.ChineseCLIPFeatureExtractor

<source>

( *args**kwargs )

ChineseCLIPProcessor

class transformers.ChineseCLIPProcessor

<source>

( image_processor = Nonetokenizer = None**kwargs )

Parameters

Constructs a Chinese-CLIP processor which wraps a Chinese-CLIP image processor and a Chinese-CLIP tokenizer into a single processor.

ChineseCLIPProcessor offers all the functionalities of ChineseCLIPImageProcessor and BertTokenizerFast. See the __call__() and decode() for more information.

batch_decode

<source>

( *args**kwargs )

This method forwards all its arguments to BertTokenizerFastโ€™s batch_decode(). Please refer to the docstring of this method for more information.

decode

<source>

( *args**kwargs )

This method forwards all its arguments to BertTokenizerFastโ€™s decode(). Please refer to the docstring of this method for more information.

ChineseCLIPModel

class transformers.ChineseCLIPModel

<source>

( config: ChineseCLIPConfig )

Parameters

  • config (ChineseCLIPConfig) โ€” Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

This model is a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

<source>

( input_ids: typing.Optional[torch.LongTensor] = Nonepixel_values: typing.Optional[torch.FloatTensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Nonetoken_type_ids: typing.Optional[torch.Tensor] = Noneposition_ids: typing.Optional[torch.LongTensor] = Nonereturn_loss: typing.Optional[bool] = Noneoutput_attentions: typing.Optional[bool] = Noneoutput_hidden_states: typing.Optional[bool] = Nonereturn_dict: typing.Optional[bool] = None ) โ†’ transformers.models.chinese_clip.modeling_chinese_clip.ChineseCLIPOutput or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) โ€” Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide it.

    Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) โ€” Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,

    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) โ€” Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,

    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) โ€” Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)) โ€” Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using AutoImageProcessor. See ChineseCLIPImageProcessor.call() for details.

  • return_loss (bool, optional) โ€” Whether or not to return the contrastive loss.

  • output_attentions (bool, optional) โ€” Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

  • output_hidden_states (bool, optional) โ€” Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.

  • return_dict (bool, optional) โ€” Whether or not to return a ModelOutput instead of a plain tuple.

Returns

transformers.models.chinese_clip.modeling_chinese_clip.ChineseCLIPOutput or tuple(torch.FloatTensor)

A transformers.models.chinese_clip.modeling_chinese_clip.ChineseCLIPOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (<class 'transformers.models.chinese_clip.configuration_chinese_clip.ChineseCLIPConfig'>) and inputs.

  • loss (torch.FloatTensor of shape (1,), optional, returned when return_loss is True) โ€” Contrastive loss for image-text similarity.

  • logits_per_image:(torch.FloatTensor of shape (image_batch_size, text_batch_size)) โ€” The scaled dot product scores between image_embeds and text_embeds. This represents the image-text similarity scores.

  • logits_per_text:(torch.FloatTensor of shape (text_batch_size, image_batch_size)) โ€” The scaled dot product scores between text_embeds and image_embeds. This represents the text-image similarity scores.

  • text_embeds(torch.FloatTensor of shape (batch_size, output_dim) โ€” The text embeddings obtained by applying the projection layer to the pooled output of ChineseCLIPTextModel.

  • image_embeds(torch.FloatTensor of shape (batch_size, output_dim) โ€” The image embeddings obtained by applying the projection layer to the pooled output of ChineseCLIPVisionModel.

  • text_model_output(BaseModelOutputWithPoolingAndCrossAttentions): The output of the ChineseCLIPTextModel.

  • vision_model_output(BaseModelOutputWithPoolingAndCrossAttentions): The output of the ChineseCLIPVisionModel.

The ChineseCLIPModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Examples:

Copied

>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, ChineseCLIPModel

>>> model = ChineseCLIPModel.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")
>>> processor = AutoProcessor.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")

>>> url = "https://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/pokemon.jpeg"
>>> image = Image.open(requests.get(url, stream=True).raw)

>>> inputs = processor(text=["ๆฐๅฐผ้พŸ", "ๅฆ™่›™็งๅญ", "ๅฐ็ซ้พ™", "็šฎๅกไธ˜"], images=image, return_tensors="pt", padding=True)

>>> outputs = model(**inputs)
>>> logits_per_image = outputs.logits_per_image  # this is the image-text similarity score
>>> probs = logits_per_image.softmax(dim=1)  # we can take the softmax to get the label probabilities

get_text_features

<source>

( input_ids: typing.Optional[torch.Tensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Nonetoken_type_ids: typing.Optional[torch.Tensor] = Noneposition_ids: typing.Optional[torch.Tensor] = Noneoutput_attentions: typing.Optional[bool] = Noneoutput_hidden_states: typing.Optional[bool] = Nonereturn_dict: typing.Optional[bool] = None ) โ†’ text_features (torch.FloatTensor of shape (batch_size, output_dim)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) โ€” Indices of input sequence tokens in the vocabulary.

    Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) โ€” Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,

    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) โ€” Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,

    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) โ€” Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) โ€” Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

    • 1 indicates the head is not masked,

    • 0 indicates the head is masked.

  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) โ€” Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the modelโ€™s internal embedding lookup matrix.

  • output_attentions (bool, optional) โ€” Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

  • output_hidden_states (bool, optional) โ€” Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.

  • return_dict (bool, optional) โ€” Whether or not to return a ModelOutput instead of a plain tuple.

Returns

text_features (torch.FloatTensor of shape (batch_size, output_dim)

The text embeddings obtained by applying the projection layer to the final [CLS] hidden state of Text-Transformer.

The ChineseCLIPModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Examples:

Copied

>>> from transformers import AutoTokenizer, ChineseCLIPModel

>>> model = ChineseCLIPModel.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")
>>> tokenizer = AutoTokenizer.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")

>>> inputs = tokenizer(["ๆฐๅฐผ้พŸ", "ๅฆ™่›™็งๅญ", "ๅฐ็ซ้พ™", "็šฎๅกไธ˜"], padding=True, return_tensors="pt")
>>> text_features = model.get_text_features(**inputs)
>>> text_features = text_features / text_features.norm(p=2, dim=-1, keepdim=True)

get_image_features

<source>

( pixel_values: typing.Optional[torch.FloatTensor] = Noneoutput_attentions: typing.Optional[bool] = Noneoutput_hidden_states: typing.Optional[bool] = Nonereturn_dict: typing.Optional[bool] = None ) โ†’ image_features (torch.FloatTensor of shape (batch_size, output_dim)

Parameters

  • pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)) โ€” Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using AutoImageProcessor. See ChineseCLIPImageProcessor.call() for details.

  • output_attentions (bool, optional) โ€” Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

  • output_hidden_states (bool, optional) โ€” Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.

  • return_dict (bool, optional) โ€” Whether or not to return a ModelOutput instead of a plain tuple.

Returns

image_features (torch.FloatTensor of shape (batch_size, output_dim)

The image embeddings obtained by applying the projection layer to the final [CLS] hidden state of Vision-Transformer.

The ChineseCLIPModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Examples:

Copied

>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, ChineseCLIPModel

>>> model = ChineseCLIPModel.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")
>>> processor = AutoProcessor.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")

>>> url = "https://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/pokemon.jpeg"
>>> image = Image.open(requests.get(url, stream=True).raw)

>>> inputs = processor(images=image, return_tensors="pt")

>>> image_features = model.get_image_features(**inputs)
>>> image_features = image_features / image_features.norm(p=2, dim=-1, keepdim=True)

ChineseCLIPTextModel

class transformers.ChineseCLIPTextModel

<source>

( configadd_pooling_layer = True )

Parameters

  • config (ChineseCLIPConfig) โ€” Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The text model from CHINESE_CLIP without any head or projection on top. This model is a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of cross-attention is added between the self-attention layers, following the architecture described in Attention is all you need by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.

To behave as an decoder the model needs to be initialized with the is_decoder argument of the configuration set to True. To be used in a Seq2Seq model, the model needs to initialized with both is_decoder argument and add_cross_attention set to True; an encoder_hidden_states is then expected as an input to the forward pass.

forward

<source>

( input_ids: typing.Optional[torch.Tensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Nonetoken_type_ids: typing.Optional[torch.Tensor] = Noneposition_ids: typing.Optional[torch.Tensor] = Nonehead_mask: typing.Optional[torch.Tensor] = Noneinputs_embeds: typing.Optional[torch.Tensor] = Noneencoder_hidden_states: typing.Optional[torch.Tensor] = Noneencoder_attention_mask: typing.Optional[torch.Tensor] = Nonepast_key_values: typing.Optional[typing.List[torch.FloatTensor]] = Noneuse_cache: typing.Optional[bool] = Noneoutput_attentions: typing.Optional[bool] = Noneoutput_hidden_states: typing.Optional[bool] = Nonereturn_dict: typing.Optional[bool] = None ) โ†’ transformers.modeling_outputs.BaseModelOutputWithPoolingAndCrossAttentions or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) โ€” Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide it.

    Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional) โ€” Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,

    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) โ€” Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,

    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) โ€” Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)) โ€” Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using AutoImageProcessor. See ChineseCLIPImageProcessor.call() for details.

  • return_loss (bool, optional) โ€” Whether or not to return the contrastive loss.

  • output_attentions (bool, optional) โ€” Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

  • output_hidden_states (bool, optional) โ€” Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.

  • return_dict (bool, optional) โ€” Whether or not to return a ModelOutput instead of a plain tuple.

  • encoder_hidden_states (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) โ€” Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if the model is configured as a decoder.

  • encoder_attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) โ€” Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in the cross-attention if the model is configured as a decoder. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,

    • 0 for tokens that are masked.

  • past_key_values (tuple(tuple(torch.FloatTensor)) of length config.n_layers with each tuple having 4 tensors of shape (batch_size, num_heads, sequence_length - 1, embed_size_per_head)) โ€” Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.

    If past_key_values are used, the user can optionally input only the last decoder_input_ids (those that donโ€™t have their past key value states given to this model) of shape (batch_size, 1) instead of all decoder_input_ids of shape (batch_size, sequence_length).

  • use_cache (bool, optional) โ€” If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).

Returns

transformers.modeling_outputs.BaseModelOutputWithPoolingAndCrossAttentions or tuple(torch.FloatTensor)

A transformers.modeling_outputs.BaseModelOutputWithPoolingAndCrossAttentions or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (ChineseCLIPConfig) and inputs.

  • last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) โ€” Sequence of hidden-states at the output of the last layer of the model.

  • pooler_output (torch.FloatTensor of shape (batch_size, hidden_size)) โ€” Last layer hidden-state of the first token of the sequence (classification token) after further processing through the layers used for the auxiliary pretraining task. E.g. for BERT-family of models, this returns the classification token after processing through a linear layer and a tanh activation function. The linear layer weights are trained from the next sentence prediction (classification) objective during pretraining.

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) โ€” Tuple of torch.FloatTensor (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) โ€” Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

  • cross_attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True and config.add_cross_attention=True is passed or when config.output_attentions=True) โ€” Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights of the decoderโ€™s cross-attention layer, after the attention softmax, used to compute the weighted average in the cross-attention heads.

  • past_key_values (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) โ€” Tuple of tuple(torch.FloatTensor) of length config.n_layers, with each tuple having 2 tensors of shape (batch_size, num_heads, sequence_length, embed_size_per_head)) and optionally if config.is_encoder_decoder=True 2 additional tensors of shape (batch_size, num_heads, encoder_sequence_length, embed_size_per_head).

    Contains pre-computed hidden-states (key and values in the self-attention blocks and optionally if config.is_encoder_decoder=True in the cross-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

The ChineseCLIPTextModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

Copied

>>> from transformers import AutoTokenizer, ChineseCLIPTextModel
>>> import torch

>>> tokenizer = AutoTokenizer.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")
>>> model = ChineseCLIPTextModel.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")

>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs)

>>> last_hidden_states = outputs.last_hidden_state

ChineseCLIPVisionModel

class transformers.ChineseCLIPVisionModel

<source>

( config: ChineseCLIPVisionConfig )

Parameters

  • config (ChineseCLIPConfig) โ€” Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The vision model from CHINESE_CLIP without any head or projection on top. This model is a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

<source>

( pixel_values: typing.Optional[torch.FloatTensor] = Noneoutput_attentions: typing.Optional[bool] = Noneoutput_hidden_states: typing.Optional[bool] = Nonereturn_dict: typing.Optional[bool] = None ) โ†’ transformers.modeling_outputs.BaseModelOutputWithPooling or tuple(torch.FloatTensor)

Parameters

  • pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)) โ€” Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using AutoImageProcessor. See ChineseCLIPImageProcessor.call() for details.

  • output_attentions (bool, optional) โ€” Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

  • output_hidden_states (bool, optional) โ€” Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.

  • return_dict (bool, optional) โ€” Whether or not to return a ModelOutput instead of a plain tuple.

Returns

transformers.modeling_outputs.BaseModelOutputWithPooling or tuple(torch.FloatTensor)

A transformers.modeling_outputs.BaseModelOutputWithPooling or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (<class 'transformers.models.chinese_clip.configuration_chinese_clip.ChineseCLIPVisionConfig'>) and inputs.

  • last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) โ€” Sequence of hidden-states at the output of the last layer of the model.

  • pooler_output (torch.FloatTensor of shape (batch_size, hidden_size)) โ€” Last layer hidden-state of the first token of the sequence (classification token) after further processing through the layers used for the auxiliary pretraining task. E.g. for BERT-family of models, this returns the classification token after processing through a linear layer and a tanh activation function. The linear layer weights are trained from the next sentence prediction (classification) objective during pretraining.

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) โ€” Tuple of torch.FloatTensor (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) โ€” Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The ChineseCLIPVisionModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Examples:

Copied

>>> from PIL import Image
>>> import requests
>>> from transformers import CLIPProcessor, ChineseCLIPVisionModel

>>> model = ChineseCLIPVisionModel.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")
>>> processor = CLIPProcessor.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16")

>>> url = "https://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/pokemon.jpeg"
>>> image = Image.open(requests.get(url, stream=True).raw)

>>> inputs = processor(images=image, return_tensors="pt")

>>> outputs = model(**inputs)
>>> last_hidden_state = outputs.last_hidden_state
>>> pooled_output = outputs.pooler_output  # pooled CLS states

Last updated