Open-Foundation-Models/PolyNorm_1B

1# coding=utf-82# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.3#4# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX5# and OPT implementations in this library. It has been modified from its6# original forms to accommodate minor architectural differences compared7# to GPT-NeoX and OPT used by the Meta AI team that trained the model.8#9# Licensed under the Apache License, Version 2.0 (the "License");10# you may not use this file except in compliance with the License.11# You may obtain a copy of the License at12#13#     http://www.apache.org/licenses/LICENSE-2.014#15# Unless required by applicable law or agreed to in writing, software16# distributed under the License is distributed on an "AS IS" BASIS,17# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.18# See the License for the specific language governing permissions and19# limitations under the License.20""" LLaMA model configuration"""21 22from transformers.configuration_utils import PretrainedConfig23from transformers.utils import logging24 25logger = logging.get_logger(__name__)26 27LLAMA_PRETRAINED_CONFIG_ARCHIVE_MAP = {}28 29 30class PolyLlamaConfig(PretrainedConfig):31    r"""32    This is the configuration class to store the configuration of a [`LlamaModel`]. It is used to instantiate an LLaMA33    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the34    defaults will yield a similar configuration to that of the LLaMA-7B.35 36    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the37    documentation from [`PretrainedConfig`] for more information.38 39 40    Args:41        vocab_size (`int`, *optional*, defaults to 32000):42            Vocabulary size of the LLaMA model. Defines the number of different tokens that can be represented by the43            `inputs_ids` passed when calling [`LlamaModel`]44        hidden_size (`int`, *optional*, defaults to 4096):45            Dimension of the hidden representations.46        intermediate_size (`int`, *optional*, defaults to 11008):47            Dimension of the MLP representations.48        num_hidden_layers (`int`, *optional*, defaults to 32):49            Number of hidden layers in the Transformer encoder.50        num_attention_heads (`int`, *optional*, defaults to 32):51            Number of attention heads for each attention layer in the Transformer encoder.52        num_key_value_heads (`int`, *optional*):53            This is the number of key_value heads that should be used to implement Grouped Query Attention. If54            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if55            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When56            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed57            by meanpooling all the original heads within that group. For more details checkout [this58            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to59            `num_attention_heads`.60        pretraining_tp (`int`, *optional*, defaults to `1`):61            Experimental feature. Tensor parallelism rank used during pretraining. Please refer to [this62            document](https://huggingface.co/docs/transformers/parallelism) to understand more about it. This value is63            necessary to ensure exact reproducibility of the pretraining results. Please refer to [this64            issue](https://github.com/pytorch/pytorch/issues/76232).65        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):66            The non-linear activation function (function or string) in the decoder.67        max_position_embeddings (`int`, *optional*, defaults to 2048):68            The maximum sequence length that this model might ever be used with. Llama 1 supports up to 2048 tokens,69            Llama 2 up to 4096, CodeLlama up to 16384.70        initializer_range (`float`, *optional*, defaults to 0.02):71            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.72        rms_norm_eps (`float`, *optional*, defaults to 1e-12):73            The epsilon used by the rms normalization layers.74        use_cache (`bool`, *optional*, defaults to `True`):75            Whether or not the model should return the last key/values attentions (not used by all models). Only76            relevant if `config.is_decoder=True`.77        tie_word_embeddings(`bool`, *optional*, defaults to `False`):78            Whether to tie weight embeddings79        rope_theta (`float`, *optional*, defaults to 10000.0):80            The base period of the RoPE embeddings.81        rope_scaling (`Dict`, *optional*):82            Dictionary containing the scaling configuration for the RoPE embeddings. Currently supports two scaling83            strategies: linear and dynamic. Their scaling factor must be an float greater than 1. The expected format84            is `{"type": strategy name, "factor": scaling factor}`. When using this flag, don't update85            `max_position_embeddings` to the expected new maximum. See the following thread for more information on how86            these scaling strategies behave:87            https://www.reddit.com/r/LocalLLaMA/comments/14mrgpr/dynamically_scaled_rope_further_increases/. This is an88            experimental feature, subject to breaking API changes in future versions.89        attention_bias (`bool`, defaults to `False`):90            Whether to use a bias in the query, key, value and output projection layers during self-attention.91 92        Example:93 94    ```python95    >>> from transformers import LlamaModel, LlamaConfig96 97    >>> # Initializing a LLaMA llama-7b style configuration98    >>> configuration = LlamaConfig()99 100    >>> # Initializing a model from the llama-7b style configuration101    >>> model = LlamaModel(configuration)102 103    >>> # Accessing the model configuration104    >>> configuration = model.config105    ```"""106    model_type = "polyllama"107    keys_to_ignore_at_inference = ["past_key_values"]108 109    def __init__(110        self,111        vocab_size=32000,112        hidden_size=4096,113        intermediate_size=11008,114        num_hidden_layers=32,115        num_attention_heads=32,116        num_key_value_heads=None,117        hidden_act="silu",118        max_position_embeddings=2048,119        initializer_range=0.02,120        rms_norm_eps=1e-6,121        use_cache=True,122        pad_token_id=None,123        bos_token_id=1,124        eos_token_id=2,125        pretraining_tp=1,126        tie_word_embeddings=False,127        rope_theta=10000.0,128        rope_scaling=None,129        attention_bias=False,130        **kwargs,131    ):132        self.vocab_size = vocab_size133        self.max_position_embeddings = max_position_embeddings134        self.hidden_size = hidden_size135        self.intermediate_size = intermediate_size136        self.num_hidden_layers = num_hidden_layers137        self.num_attention_heads = num_attention_heads138 139        # for backward compatibility140        if num_key_value_heads is None:141            num_key_value_heads = num_attention_heads142 143        self.num_key_value_heads = num_key_value_heads144        self.hidden_act = hidden_act145        self.initializer_range = initializer_range146        self.rms_norm_eps = rms_norm_eps147        self.pretraining_tp = pretraining_tp148        self.use_cache = use_cache149        self.rope_theta = rope_theta150        self.rope_scaling = rope_scaling151        self._rope_scaling_validation()152        self.attention_bias = attention_bias153 154        super().__init__(155            pad_token_id=pad_token_id,156            bos_token_id=bos_token_id,157            eos_token_id=eos_token_id,158            tie_word_embeddings=tie_word_embeddings,159            **kwargs,160        )161 162    def _rope_scaling_validation(self):163        """164        Validate the `rope_scaling` configuration.165        """166        if self.rope_scaling is None:167            return168 169        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 2:170            raise ValueError(171                "`rope_scaling` must be a dictionary with with two fields, `type` and `factor`, "172                f"got {self.rope_scaling}"173            )174        rope_scaling_type = self.rope_scaling.get("type", None)175        rope_scaling_factor = self.rope_scaling.get("factor", None)176        if rope_scaling_type is None or rope_scaling_type not in ["linear", "dynamic"]:177            raise ValueError(178                f"`rope_scaling`'s type field must be one of ['linear', 'dynamic'], got {rope_scaling_type}"179            )180        if rope_scaling_factor is None or not isinstance(rope_scaling_factor, float) or rope_scaling_factor <= 1.0:181            raise ValueError(f"`rope_scaling`'s factor field must be an float > 1, got {rope_scaling_factor}")