TIGER-Lab/VLM2Vec-Full

1# coding=utf-82# Copyright 2024 Microsoft and the HuggingFace Inc. team. All rights reserved.3#4# Licensed under the Apache License, Version 2.0 (the "License");5# you may not use this file except in compliance with the License.6# You may obtain a copy of the License at7#8#     http://www.apache.org/licenses/LICENSE-2.09#10# Unless required by applicable law or agreed to in writing, software11# distributed under the License is distributed on an "AS IS" BASIS,12# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.13# See the License for the specific language governing permissions and14# limitations under the License.15 16""" Phi-3-V model configuration"""17 18 19from transformers.configuration_utils import PretrainedConfig20from transformers.utils import logging21 22 23logger = logging.get_logger(__name__)24 25PHI3V_PRETRAINED_CONFIG_ARCHIVE_MAP = {26    "microsoft/Phi-3-vision-128k-instruct": "https://huggingface.co/microsoft/Phi-3-vision-128k-instruct/resolve/main/config.json",27    "microsoft/Phi-3.5-vision-instruct": "https://huggingface.co/microsoft/Phi-3.5-vision-instruct/resolve/main/config.json",28}29 30 31class Phi3VConfig(PretrainedConfig):32    r"""33    This is the configuration class to store the configuration of a [`Phi3VModel`]. It is used to instantiate a Phi-334    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the35    defaults will yield a similar configuration to that of the36    [microsoft/Phi-3-vision-128k-instruct](https://huggingface.co/microsoft/Phi-3-vision-128k-instruct).37 38    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the39    documentation from [`PretrainedConfig`] for more information.40 41    Args:42        vocab_size (`int`, *optional*, defaults to 32064):43            Vocabulary size of the Phi-3-V model. Defines the number of different tokens that can be represented by the44            `inputs_ids` passed when calling [`Phi3VModel`].45        hidden_size (`int`, *optional*, defaults to 3072):46            Dimension of the hidden representations.47        intermediate_size (`int`, *optional*, defaults to 8192):48            Dimension of the MLP representations.49        num_hidden_layers (`int`, *optional*, defaults to 32):50            Number of hidden layers in the Transformer decoder.51        num_attention_heads (`int`, *optional*, defaults to 32):52            Number of attention heads for each attention layer in the Transformer decoder.53        num_key_value_heads (`int`, *optional*):54            This is the number of key_value heads that should be used to implement Grouped Query Attention. If55            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if56            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When57            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed58            by meanpooling all the original heads within that group. For more details checkout [this59            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to60            `num_attention_heads`.61        resid_pdrop (`float`, *optional*, defaults to 0.0):62            Dropout probability for mlp outputs.63        embd_pdrop (`int`, *optional*, defaults to 0.0):64            The dropout ratio for the embeddings.65        attention_dropout (`float`, *optional*, defaults to 0.0):66            The dropout ratio after computing the attention scores.67        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):68            The non-linear activation function (function or string) in the decoder.69        max_position_embeddings (`int`, *optional*, defaults to 4096):70            The maximum sequence length that this model might ever be used with.71        original_max_position_embeddings (`int`, *optional*, defaults to 4096):72            The maximum sequence length that this model was trained with. This is used to determine the size of the73            original RoPE embeddings when using long scaling.74        initializer_range (`float`, *optional*, defaults to 0.02):75            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.76        rms_norm_eps (`float`, *optional*, defaults to 1e-05):77            The epsilon value used for the RMSNorm.78        use_cache (`bool`, *optional*, defaults to `True`):79            Whether or not the model should return the last key/values attentions (not used by all models). Only80            relevant if `config.is_decoder=True`. Whether to tie weight embeddings or not.81        tie_word_embeddings (`bool`, *optional*, defaults to `False`):82            Whether to tie weight embeddings83        rope_theta (`float`, *optional*, defaults to 10000.0):84            The base period of the RoPE embeddings.85        rope_scaling (`dict`, *optional*):86            The scaling strategy for the RoPE embeddings. If `None`, no scaling is applied. If a dictionary, it must87            contain the following keys: `type`, `short_factor` and `long_factor`. The `type` must be either `su` or `yarn` and88            the `short_factor` and `long_factor` must be lists of numbers with the same length as the hidden size89            divided by the number of attention heads divided by 2.90        bos_token_id (`int`, *optional*, defaults to 1):91            The id of the "beginning-of-sequence" token.92        eos_token_id (`int`, *optional*, defaults to 32000):93            The id of the "end-of-sequence" token.94        pad_token_id (`int`, *optional*, defaults to 32000):95            The id of the padding token.96        sliding_window (`int`, *optional*):97            Sliding window attention window size. If `None`, no sliding window is applied.98        embd_layer (`str`, *optional*, defaults to `"default"`):99            The embedding layer to use. Can be either `"default"` or `"image"`. "default" uses the standard embedding for text. 100 101    Example:102 103    ```python104    >>> from transformers import Phi3VModel, Phi3VConfig105 106    >>> # Initializing a Phi-3-V style configuration107    >>> configuration = Phi3Config.from_pretrained("microsoft/Phi-3-vision-128k-instruct")108 109    >>> # Initializing a model from the configuration110    >>> model = Phi3VModel(configuration)111 112    >>> # Accessing the model configuration113    >>> configuration = model.config114    ```"""115 116    model_type = "phi3_v"117    keys_to_ignore_at_inference = ["past_key_values"]118 119    def __init__(120        self,121        vocab_size=32064,122        hidden_size=3072,123        intermediate_size=8192,124        num_hidden_layers=32,125        num_attention_heads=32,126        num_key_value_heads=None,127        resid_pdrop=0.0,128        embd_pdrop=0.0,129        attention_dropout=0.0,130        hidden_act="silu",131        max_position_embeddings=4096,132        original_max_position_embeddings=4096,133        initializer_range=0.02,134        rms_norm_eps=1e-5,135        use_cache=True,136        tie_word_embeddings=False,137        rope_theta=10000.0,138        rope_scaling=None,139        bos_token_id=1,140        eos_token_id=32000,141        pad_token_id=32000,142        sliding_window=None,143        embd_layer: str = "default",144        **kwargs,145    ):146        self.vocab_size = vocab_size147        self.hidden_size = hidden_size148        self.intermediate_size = intermediate_size149        self.num_hidden_layers = num_hidden_layers150        self.num_attention_heads = num_attention_heads151 152        if num_key_value_heads is None:153            num_key_value_heads = num_attention_heads154 155        self.num_key_value_heads = num_key_value_heads156        self.resid_pdrop = resid_pdrop157        self.embd_pdrop = embd_pdrop158        self.attention_dropout = attention_dropout159        self.hidden_act = hidden_act160        self.max_position_embeddings = max_position_embeddings161        self.original_max_position_embeddings = original_max_position_embeddings162        self.initializer_range = initializer_range163        self.rms_norm_eps = rms_norm_eps164        self.use_cache = use_cache165        self.rope_theta = rope_theta166        self.rope_scaling = rope_scaling167        self._rope_scaling_validation()168        self.sliding_window = sliding_window169        self.embd_layer = embd_layer170 171 172        super().__init__(173            bos_token_id=bos_token_id,174            eos_token_id=eos_token_id,175            pad_token_id=pad_token_id,176            tie_word_embeddings=tie_word_embeddings,177            **kwargs,178        )179 180    def _rope_scaling_validation(self):181        """182        Validate the `rope_scaling` configuration.183        """184        if self.rope_scaling is None:185            return186 187        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 3:188            raise ValueError(189                "`rope_scaling` must be a dictionary with three fields, `type`, `short_factor` and `long_factor`, "190                f"got {self.rope_scaling}"191            )192        rope_scaling_type = self.rope_scaling.get("type", None)193        rope_scaling_short_factor = self.rope_scaling.get("short_factor", None)194        rope_scaling_long_factor = self.rope_scaling.get("long_factor", None)195        if rope_scaling_type is None or rope_scaling_type not in ["su", "yarn"]:196            raise ValueError(f"`rope_scaling`'s type field must be one of ['su', 'yarn'], got {rope_scaling_type}")197        if not (198            isinstance(rope_scaling_short_factor, list)199            and all(isinstance(x, (int, float)) for x in rope_scaling_short_factor)200        ):201            raise ValueError(202                f"`rope_scaling`'s short_factor field must be a list of numbers, got {rope_scaling_short_factor}"203            )204        if not len(rope_scaling_short_factor) == self.hidden_size // self.num_attention_heads // 2:205            raise ValueError(206                f"`rope_scaling`'s short_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_short_factor)}"207            )208        if not (209            isinstance(rope_scaling_long_factor, list)210            and all(isinstance(x, (int, float)) for x in rope_scaling_long_factor)211        ):212            raise ValueError(213                f"`rope_scaling`'s long_factor field must be a list of numbers, got {rope_scaling_long_factor}"214            )215        if not len(rope_scaling_long_factor) == self.hidden_size // self.num_attention_heads // 2:216            raise ValueError(217                f"`rope_scaling`'s long_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_long_factor)}"218            )