# coding=utf-8 # Copyright 2024 The Qwen team, Alibaba Group and the HuggingFace Inc. team. 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. """Qwen2 model configuration""" from transformers.configuration_utils import PretrainedConfig from transformers.utils import logging logger = logging.get_logger(__name__) class DarwinLMConfig(PretrainedConfig): r""" 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 = "darwinlm" 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, kv_ignore=False, dim_each_mlp={"0.mlp.down_proj": 22016, "1.mlp.down_proj": 22016, "2.mlp.down_proj": 22016, "3.mlp.down_proj": 22016, "4.mlp.down_proj": 22016, "5.mlp.down_proj": 22016, "6.mlp.down_proj": 22016, "7.mlp.down_proj": 22016, "8.mlp.down_proj": 22016, "9.mlp.down_proj": 22016, "10.mlp.down_proj": 22016, "11.mlp.down_proj": 22016, "12.mlp.down_proj": 22016, "13.mlp.down_proj": 22016, "14.mlp.down_proj": 22016, "15.mlp.down_proj": 22016, "16.mlp.down_proj": 22016, "17.mlp.down_proj": 22016, "18.mlp.down_proj": 22016, "19.mlp.down_proj": 22016, "20.mlp.down_proj": 22016, "21.mlp.down_proj": 22016, "22.mlp.down_proj": 22016, "23.mlp.down_proj": 22016, "24.mlp.down_proj": 22016, "25.mlp.down_proj": 22016, "26.mlp.down_proj": 22016, "27.mlp.down_proj": 22016, "28.mlp.down_proj": 22016, "29.mlp.down_proj": 22016, "30.mlp.down_proj": 22016, "31.mlp.down_proj": 22016, "32.mlp.down_proj": 22016, "33.mlp.down_proj": 22016, "34.mlp.down_proj": 22016, "35.mlp.down_proj": 22016, "36.mlp.down_proj": 22016, "37.mlp.down_proj": 22016, "38.mlp.down_proj": 22016, "39.mlp.down_proj": 22016, "40.mlp.down_proj": 22016, "41.mlp.down_proj": 22016, "42.mlp.down_proj": 22016, "43.mlp.down_proj": 22016, "44.mlp.down_proj": 22016, "45.mlp.down_proj": 22016, "46.mlp.down_proj": 22016, "47.mlp.down_proj": 22016,}, heads_each_attn={"0.self_attn.o_proj": 32, "1.self_attn.o_proj": 32, "2.self_attn.o_proj": 32, "3.self_attn.o_proj": 32, "4.self_attn.o_proj": 32, "5.self_attn.o_proj": 32, "6.self_attn.o_proj": 32, "7.self_attn.o_proj": 32, "8.self_attn.o_proj": 32, "9.self_attn.o_proj": 32, "10.self_attn.o_proj": 32, "11.self_attn.o_proj": 32, "12.self_attn.o_proj": 32, "13.self_attn.o_proj": 32, "14.self_attn.o_proj": 32, "15.self_attn.o_proj": 32, "16.self_attn.o_proj": 32, "17.self_attn.o_proj": 32, "18.self_attn.o_proj": 32, "19.self_attn.o_proj": 32, "20.self_attn.o_proj": 32, "21.self_attn.o_proj": 32, "22.self_attn.o_proj": 32, "23.self_attn.o_proj": 32, "24.self_attn.o_proj": 32, "25.self_attn.o_proj": 32, "26.self_attn.o_proj": 32, "27.self_attn.o_proj": 32, "28.self_attn.o_proj": 32, "29.self_attn.o_proj": 32, "30.self_attn.o_proj": 32, "31.self_attn.o_proj": 32, "32.self_attn.o_proj": 32, "33.self_attn.o_proj": 32, "34.self_attn.o_proj": 32, "35.self_attn.o_proj": 32, "36.self_attn.o_proj": 32, "37.self_attn.o_proj": 32, "38.self_attn.o_proj": 32, "39.self_attn.o_proj": 32, "40.self_attn.o_proj": 32, "41.self_attn.o_proj": 32, "42.self_attn.o_proj": 32, "43.self_attn.o_proj": 32, "44.self_attn.o_proj": 32, "45.self_attn.o_proj": 32, "46.self_attn.o_proj": 32, "47.self_attn.o_proj": 32,}, **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.kv_ignore = kv_ignore self.heads_each_attn = heads_each_attn self.dim_each_mlp = dim_each_mlp # 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 super().__init__( tie_word_embeddings=tie_word_embeddings, **kwargs, )