File size: 12,622 Bytes
5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a 5985681 067141a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 |
# coding=utf-8
# Copyright 2024 Jingze Shi and the HuggingFace Inc. team. All rights reserved.
#
# This code is based on the Wonderful Matrices paper implementation.
#
# https://arxiv.org/abs/2412.11834
#
# 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.
"""PyTorch Doge model configuration"""
from transformers.configuration_utils import PretrainedConfig
from transformers.modeling_rope_utils import rope_config_validation
class DogeConfig(PretrainedConfig):
r"""
This is the configuration class to store the configuration of a [`DogeModel`]. It is used to instantiate an Doge
model according to the specified arguments, defining the model architecture like [JingzeShi/Doge-20M](https://huggingface.co/JingzeShi/Doge-20M).
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 32768):
Vocabulary size of the Doge model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`DogeModel`]
num_channels (`int`, *optional*, defaults to 3):
Number of channels in the input image.
patch_size (`int`, *optional*, defaults to 16):
Patch size of Vision Transformer Embeddings.
hidden_size (`int`, *optional*, defaults to 1024):
Dimension of the hidden representations.
intermediate_size (`int`, *optional*, defaults to 2048):
Dimension of the MLP representations.
num_hidden_layers (`int`, *optional*, defaults to 32):
Number of hidden layers in the Transformer decoder.
hidden_bias (`bool`, *optional*, defaults to `False`):
Whether to use bias in the hidden layers.
hidden_dropout (`float`, *optional*, defaults to 0.0):
Dropout probability for each sequence transformation and state transformation module.
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 2048):
The maximum sequence length that this model might ever be used with.
rope_theta (`float`, *optional*, defaults to 10000.0):
The base period of the RoPE embeddings.
rope_scaling (`Dict`, *optional*):
Dictionary containing the scaling configuration for the RoPE embeddings.
NOTE: if you apply new rope type and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value accordingly.
Expected contents:
`rope_type` (`str`):
The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope', 'llama3'], with 'default' being the original RoPE implementation.
`factor` (`float`, *optional*):
Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings.
In most scaling types, a `factor` of x will enable the model to handle sequences of length x * original maximum pre-trained length.
`original_max_position_embeddings` (`int`, *optional*):
Used with 'dynamic', 'longrope' and 'llama3'.
The original max position embeddings used during pretraining.
`attention_factor` (`float`, *optional*):
Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention
computation.
If unspecified, it defaults to value recommended by the implementation, using the `factor` field to infer the suggested value.
`beta_fast` (`float`, *optional*):
Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear
ramp function. If unspecified, it defaults to 32.
`beta_slow` (`float`, *optional*):
Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear
ramp function. If unspecified, it defaults to 1.
`short_factor` (`List[float]`, *optional*):
Only used with 'longrope'. The scaling factor to be applied to short contexts (<`original_max_position_embeddings`).
Must be a list of numbers with the same length as the hidden size divided by the number of attention heads divided by 2
`long_factor` (`List[float]`, *optional*):
Only used with 'longrope'. The scaling factor to be applied to long contexts (<`original_max_position_embeddings`).
Must be a list of numbers with the same length as the hidden size divided by the number of attention heads divided by 2
`low_freq_factor` (`float`, *optional*):
Only used with 'llama3'. Scaling factor applied to low frequency components of the RoPE
`high_freq_factor` (`float`, *optional*):
Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE
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`.
pad_token_id (`int`, *optional*, defaults to 0):
Padding token id.
bos_token_id (`int`, *optional*, defaults to 1):
Beginning of stream token id.
eos_token_id (`int`, *optional*, defaults to 2):
End of stream token id.
tie_word_embeddings (`bool`, *optional*, defaults to `True`):
Whether to tie weight embeddings
num_attention_heads (`int`, *optional*, defaults to 8):
Number of attention heads for each attention layer in the Transformer decoder.
num_key_value_heads (`int`, *optional*, defaults to `None`):
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 `num_attention_heads`.
attention_dropout (`float`, *optional*, defaults to 0.0):
The dropout ratio for the attention probabilities.
dynamic_mask_ratio (`float`, *optional*, defaults to 0.0, range [0, 1]):
The ratio to control the proportion of the dynamic mask filled with the minimum value.
is_moe (`bool`, *optional*, defaults to `False`):
Whether to use the Cross Domain Mixture of Experts, if `True`, the MoE will inherit the MLP to initialize
num_cdmoe_experts (`int`, *optional*, defaults to 16348):
Number of Private Experts for the Cross Domain Mixture of Experts. calculation formula: :math:`\text{num_cdmoe_experts} = (32 \times \text{num_cdmoe_heads})^2`
num_cdmoe_heads (`int`, *optional*, defaults to 4):
Number of heads of Private Experts for the Cross Domain Mixture of Experts.
num_cdmoe_experts_per_head (`int`, *optional*, defaults to 8):
Number of Private Experts per head for the Cross Domain Mixture of Experts.
expert_retrieval_size (`int`, *optional*, defaults to 64):
Dimension of the Expert retrieval states for the Cross Domain Mixture of Experts.
"""
model_type = "doge"
keys_to_ignore_at_inference = ["past_key_values"]
# Default tensor parallel plan for base model `DogeModel`
base_model_tp_plan = {
"layers.*.self_attn.q_proj": "colwise",
"layers.*.self_attn.k_proj": "colwise",
"layers.*.self_attn.v_proj": "colwise",
"layers.*.self_attn.dt_proj": "colwise",
"layers.*.self_attn.o_proj": "rowwise",
"layers.*.mlp.gate_proj": "colwise",
"layers.*.mlp.up_proj": "colwise",
"layers.*.mlp.down_proj": "rowwise",
}
def __init__(
self,
vocab_size=32768,
num_channels=3,
patch_size=16,
hidden_size=1024,
intermediate_size=2048,
num_hidden_layers=32,
hidden_bias=False,
hidden_dropout=0.0,
hidden_act="silu",
max_position_embeddings=2048,
rope_theta=10000.0,
rope_scaling={
"rope_type": "dynamic",
"factor": 4.0,
"original_max_position_embeddings": 2048,
},
initializer_range=0.02,
rms_norm_eps=1e-06,
use_cache=True,
bos_token_id=0,
eos_token_id=1,
pad_token_id=2,
tie_word_embeddings=True,
num_attention_heads=8,
num_key_value_heads=None,
attention_dropout=0.0,
dynamic_mask_ratio=0.0,
is_moe=False,
num_cdmoe_experts=16348,
num_cdmoe_heads=4,
num_cdmoe_experts_per_head=8,
expert_retrieval_size=64,
**kwargs,
):
self.vocab_size = vocab_size
self.num_channels = num_channels
self.patch_size = patch_size
self.hidden_size = hidden_size
self.intermediate_size = intermediate_size
self.num_hidden_layers = num_hidden_layers
self.hidden_bias = hidden_bias
self.hidden_dropout = hidden_dropout
self.hidden_act = hidden_act
self.max_position_embeddings = max_position_embeddings
self.rope_theta = rope_theta
self.rope_scaling = rope_scaling
self.initializer_range = initializer_range
self.rms_norm_eps = rms_norm_eps
self.use_cache = use_cache
self.bos_token_id = bos_token_id
self.eos_token_id = eos_token_id
self.pad_token_id = pad_token_id
self.tie_word_embeddings = tie_word_embeddings
self.num_attention_heads = num_attention_heads
self.num_key_value_heads = num_key_value_heads
self.attention_dropout = attention_dropout
self.dynamic_mask_ratio = dynamic_mask_ratio
self.is_moe = is_moe
self.num_cdmoe_experts = num_cdmoe_experts
self.num_cdmoe_heads = num_cdmoe_heads
self.num_cdmoe_experts_per_head = num_cdmoe_experts_per_head
self.expert_retrieval_size = expert_retrieval_size
# Validate the correctness of rotary position embeddings parameters
# BC: if there is a 'type' field, copy it it to 'rope_type'.
if self.rope_scaling is not None and "type" in self.rope_scaling:
self.rope_scaling["rope_type"] = self.rope_scaling["type"]
rope_config_validation(self)
# for backward compatibility
if num_key_value_heads is None:
self.num_key_value_heads = num_attention_heads
super().__init__(
bos_token_id=bos_token_id,
eos_token_id=eos_token_id,
pad_token_id=pad_token_id,
tie_word_embeddings=tie_word_embeddings,
**kwargs,
)
__all__ = ["DogeConfig"]
|