You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

207 lines
11 KiB

# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# This file was automatically generated from src/transformers/models/doge/modular_doge.py.
# Do NOT edit this file manually as any edits will be overwritten by the generation of
# the file from the modular. If any change should be done, please apply the change to the
# modular_doge.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# Copyright 2025 Jingze Shi and the HuggingFace Inc. team. All rights reserved.
#
# The Doge family of small language models is trained by SmallDoge Team.
#
# 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.
from ...configuration_utils import PreTrainedConfig
from ...modeling_rope_utils import RopeParameters
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 [SmallDoge/Doge-320M](https://huggingface.co/SmallDoge/Doge-320M).
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 Doge2 model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`DogeModel`]
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_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.
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.
max_position_embeddings (`int`, *optional*, defaults to 2048):
The maximum sequence length that this model might ever be used with.
rope_parameters (`RopeParameters`, *optional*):
Dictionary containing the configuration parameters for the RoPE embeddings. The dictionary should contain
a value for `rope_theta` and optionally parameters used for scaling in case you want to use RoPE
with longer `max_position_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*):
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://huggingface.co/papers/2305.13245).
If it is not specified, will default to `num_attention_heads`.
attention_bias (`bool`, defaults to `False`, *optional*, defaults to `False`):
Whether to use a bias in the query, key, value and output projection layers during self-attention.
attention_dropout (`float`, *optional*, defaults to 0.0):
The dropout ratio for the attention probabilities.
mlp_bias (`bool`, *optional*, defaults to `False`):
Whether to use a bias in up_proj, down_proj and gate_proj layers in the MLP layers.
sliding_window (`int`, *optional*):
Sliding window attention window size. If not specified, will default to `None`.
keep_window_size (`int`, *optional*, defaults to 2048):
The window size of tokens that are not dynamically masked, and dynamic masking is only performed when the sequence length exceeds this 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_experts (`int`, *optional*, defaults to 16384):
Number of routed experts in the model. This is only used when `is_moe=True`.
num_experts_per_tok (`int`, *optional*, defaults to 64):
Number of selected experts to route per-token.
norm_topk_prob (`bool`, *optional*, defaults to `False`):
Whether to normalize the topk probabilities.
output_router_logits (`bool`, *optional*, defaults to `False`):
Whether or not the router logits should be returned by the model. Enabling this will also
allow the model to output the auxiliary loss, including load balancing loss and router z-loss.
router_aux_loss_coef (`float`, *optional*, defaults to 0.001):
The aux loss factor for the total loss.
pad_token_id (`int`, *optional*):
Padding token id.
bos_token_id (`int`, *optional*):
Beginning of stream token id.
eos_token_id (`int`, *optional*):
End of stream token id.
```python
>>> from transformers import DogeConfig, DogeModel
>>> # Initializing a Doge-320M style configuration
>>> configuration = DogeConfig()
>>> # Initializing a model from the Doge-320M style configuration
>>> model = DogeModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
```"""
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": "rowwise",
"layers.*.self_attn.o_proj": "rowwise",
"layers.*.mlp.gate_proj": "colwise",
"layers.*.mlp.up_proj": "colwise",
"layers.*.mlp.down_proj": "rowwise",
"layers.*.mlp.router_gate": "colwise_gather_output",
"layers.*.mlp.down_embed": "rowwise_split_input",
"layers.*.mlp.up_embed": "rowwise_split_input",
}
base_model_pp_plan = {
"embed_tokens": (["input_ids"], ["inputs_embeds"]),
"layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
"norm": (["hidden_states"], ["hidden_states"]),
}
def __init__(
self,
vocab_size: int | None = 32768,
hidden_size: int | None = 1024,
intermediate_size: int | None = 2048,
num_hidden_layers: int | None = 32,
hidden_dropout: float | None = 0.0,
hidden_act: str | None = "silu",
initializer_range: float | None = 0.02,
rms_norm_eps: int | None = 1e-06,
use_cache: bool | None = True,
tie_word_embeddings: bool | None = False,
max_position_embeddings: int | None = 2048,
rope_parameters: RopeParameters | dict[str, RopeParameters] | None = None,
num_attention_heads: int | None = 8,
num_key_value_heads: int | None = None,
attention_bias: bool | None = False,
attention_dropout: float | None = 0.0,
mlp_bias: bool | None = False,
sliding_window: int | None = None,
keep_window_size: int | None = 2048,
is_moe: bool | None = False,
num_experts: int | None = 16384,
num_experts_per_tok: int | None = 64,
norm_topk_prob: bool | None = False,
output_router_logits: bool | None = False,
router_aux_loss_coef: float | None = 0.001,
pad_token_id: int | None = None,
bos_token_id: int | None = None,
eos_token_id: int | None = None,
**kwargs,
):
self.vocab_size = vocab_size
self.hidden_size = hidden_size
self.intermediate_size = intermediate_size
self.num_hidden_layers = num_hidden_layers
self.hidden_dropout = hidden_dropout
self.hidden_act = hidden_act
self.initializer_range = initializer_range
self.rms_norm_eps = rms_norm_eps
self.use_cache = use_cache
self.max_position_embeddings = max_position_embeddings
self.num_attention_heads = num_attention_heads
self.num_key_value_heads = num_key_value_heads
self.attention_bias = attention_bias
self.attention_dropout = attention_dropout
self.mlp_bias = mlp_bias
self.sliding_window = sliding_window
self.keep_window_size = keep_window_size
self.is_moe = is_moe
self.num_experts = num_experts
self.num_experts_per_tok = num_experts_per_tok
self.norm_topk_prob = norm_topk_prob
self.output_router_logits = output_router_logits
self.router_aux_loss_coef = router_aux_loss_coef
self.tie_word_embeddings = tie_word_embeddings
self.pad_token_id = pad_token_id
self.bos_token_id = bos_token_id
self.eos_token_id = eos_token_id
self.rope_parameters = rope_parameters
# for backward compatibility
if num_key_value_heads is None:
self.num_key_value_heads = num_attention_heads
super().__init__(**kwargs)
__all__ = ["DogeConfig"]