# Copyright 2025 The EasyDeL Author @erfanzar (Erfan Zare Chavoshi).
#
# 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
#
# https://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.
import typing as tp
from eformer.common_types import (
EMPTY,
MODE_TRAIN,
TP,
ColumnWise,
DynamicShardingAxes,
Replicated,
RowWise,
)
from easydel.infra.base_module import EasyDeLBaseConfig
from easydel.infra.etils import EasyDeLGradientCheckPointers
from easydel.infra.factory import register_config
from easydel.infra.utils import AttnMaskDetail, AttnMaskType
from easydel.layers.moe.utils import get_moe_partition_spec
[docs]class ExpertTensorParallel(DynamicShardingAxes):
"""Expert Tensor Parallelism (EPxTP) sharding axes."""
axes: tp.ClassVar = [TP, EMPTY, EMPTY]
mode: tp.ClassVar = MODE_TRAIN
[docs]@register_config("arctic")
class ArcticConfig(EasyDeLBaseConfig):
"""
Configuration objects inherit from [`EasyDeLBaseConfig`] and can be used to control the model outputs. Read
the documentation from [`EasyDeLBaseConfig`] for more information.
Args:
vocab_size (`int`, *optional*, defaults to 32000):
Vocabulary size of the ARCTIC model. Defines the number of different tokens that can be represented by the
`inputs_ids` passed to the forward method.
hidden_size (`int`, *optional*, defaults to 4096):
Dimensionality of the encoder layers and the pooler layer.
intermediate_size (`int`, *optional*, defaults to 14336):
Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
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*):
Number of key and value heads for each attention layer in the Transformer encoder. Will default to
`num_attention_heads` if not set.
hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
The non-linear activation function (function or string) to use in the encoder and pooler. If string,
`"gelu"`, `"relu"`, `"swish"` and `"gelu_new"` are supported.
max_position_embeddings (`int`, *optional*, defaults to 4096):
The maximum sequence length that this model might ever be used with. Typically set this to something large
just in case (e.g., 2048 or 4096).
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-5):
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*):
The index of the padding token in the vocabulary. The default value (`0`) is the same as for GPT2.
bos_token_id (`int`, *optional*):
The index of the beginning of sequence token in the vocabulary. The default value (`1`) is the same as for
GPT2.
eos_token_id (`int`, *optional*):
The index of the end of sequence token in the vocabulary. The default value (`2`) is the same as for GPT2.
tie_word_embeddings (`bool`, *optional*, defaults to `False`):
Whether to tie the weights of the input embeddings and the output embeddings.
rope_theta (`float`, *optional*, defaults to 1e6):
The theta value to use for rotary position embeddings.
sliding_window (`int`, *optional*):
The sliding window size to use for attention. If not specified, no sliding window attention is used.
attention_dropout (`float`, *optional*, defaults to 0.0):
The dropout ratio for the attention probabilities.
num_experts_per_tok (`int`, *optional*, defaults to 1):
The number of experts per token for mixture of experts.
num_local_experts (`int`, *optional*, defaults to 8):
The number of local experts for mixture of experts.
router_aux_loss_coef (`float`, *optional*, defaults to 0.001):
The auxiliary loss coefficient for the router.
moe_layer_frequency (`int`, *optional*, defaults to 2):
The frequency of MoE layers.
parallel_attn_mlp_res (`bool`, *optional*, defaults to `False`):
Whether to parallelize attention and MLP residual connections.
moe_train_capacity_factor (`float`, *optional*, defaults to 1):
The capacity factor for MoE layers during training.
moe_eval_capacity_factor (`float`, *optional*, defaults to 1):
The capacity factor for MoE layers during evaluation.
enable_expert_tensor_parallelism (`bool`, *optional*, defaults to `False`):
Whether to enable expert tensor parallelism.
moe_min_capacity (`int`, *optional*, defaults to 0):
The minimum capacity for MoE layers.
moe_token_dropping (`bool`, *optional*, defaults to `True`):
Whether to drop tokens in MoE layers.
quantization (`str`, *optional*):
The quantization configuration.
gradient_checkpointing (`str`, *optional*, defaults to `"nothing_saveable"`):
The gradient checkpointing configuration.
use_scan_mlp (`bool`, *optional*, defaults to `False`):
Whether to use scan for MLP.
scan_mlp_chunk_size (`int`, *optional*, defaults to 1024):
The chunk size for scan MLP.
bits (`int`, *optional*):
The number of bits.
rope_scaling (`tp.Dict[str, tp.Union[str, float]]`, *optional*):
The rope scaling configuration.
"""
model_type: str = "arctic"
def __init__(
self,
vocab_size=32000,
hidden_size=4096,
intermediate_size=14336,
num_hidden_layers=32,
num_attention_heads=32,
num_key_value_heads=None,
hidden_act="silu",
max_position_embeddings=4096,
initializer_range=0.02,
rms_norm_eps=1e-5,
use_cache=True,
pad_token_id=None,
bos_token_id=1,
eos_token_id=2,
tie_word_embeddings=False,
rope_theta=1e6,
sliding_window=None,
attention_dropout=0.0,
num_experts_per_tok=1,
num_local_experts=8,
router_aux_loss_coef=0.001,
moe_layer_frequency=2,
parallel_attn_mlp_res=False,
moe_train_capacity_factor=1,
moe_eval_capacity_factor=1,
enable_expert_tensor_parallelism=False,
moe_min_capacity=0,
moe_token_dropping=True,
quantization=None,
gradient_checkpointing: EasyDeLGradientCheckPointers = EasyDeLGradientCheckPointers.NONE,
use_scan_mlp: bool = False,
scan_mlp_chunk_size: int = 1024,
layer_types: list[str] | None = None,
bits: int | None = None,
rope_scaling: dict[str, str | float] | None = None,
**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.sliding_window = sliding_window
# 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
self.num_experts_per_tok = num_experts_per_tok
self.num_local_experts = num_local_experts
self.router_aux_loss_coef = router_aux_loss_coef
self.moe_layer_frequency = moe_layer_frequency
self.moe_train_capacity_factor = moe_train_capacity_factor
self.moe_eval_capacity_factor = moe_eval_capacity_factor
self.enable_expert_tensor_parallelism = enable_expert_tensor_parallelism
self.moe_min_capacity = moe_min_capacity
self.moe_token_dropping = moe_token_dropping
self.parallel_attn_mlp_res = parallel_attn_mlp_res
self.quantization = quantization
self.gradient_checkpointing = gradient_checkpointing
self.use_scan_mlp = use_scan_mlp
self.scan_mlp_chunk_size = scan_mlp_chunk_size
self.bits = bits
self.rope_scaling = rope_scaling
self.layer_types = layer_types
if self.layer_types is None:
self.layer_types = [
"sliding_attention" if self.sliding_window is not None else "full_attention"
for i in range(self.num_hidden_layers)
]
super().__init__(
pad_token_id=pad_token_id,
bos_token_id=bos_token_id,
eos_token_id=eos_token_id,
tie_word_embeddings=tie_word_embeddings,
**kwargs,
)
[docs] def get_partition_rules(self, *args, **kwargs):
"""
Shard parameters for Arctic model with MoE support.
"""
pmag = self.partition_manager
return (
(r"model/embed_tokens/embedding", pmag.resolve(ColumnWise)),
(r"self_attn/(q_proj|k_proj|v_proj)/kernel", pmag.resolve(ColumnWise)),
(r"self_attn/o_proj/kernel", pmag.resolve(RowWise)),
(
r"block_sparse_moe/gate/kernel",
pmag.resolve(Replicated if self.use_expert_tensor_mode else ColumnWise),
),
(
r"block_sparse_moe/experts/.*/(w1|w3)/kernel",
get_moe_partition_spec(
partition_manager=self.partition_manager,
direction="column",
tensors_are_expert=self.use_expert_tensor_mode,
is_bias=False,
fsdp_is_ep_bound=self.fsdp_is_ep_bound,
sp_is_ep_bound=self.sp_is_ep_bound,
module_view=True,
),
),
(
r"block_sparse_moe/experts/.*/w2/kernel",
get_moe_partition_spec(
partition_manager=self.partition_manager,
direction="row",
tensors_are_expert=self.use_expert_tensor_mode,
is_bias=False,
fsdp_is_ep_bound=self.fsdp_is_ep_bound,
sp_is_ep_bound=self.sp_is_ep_bound,
module_view=True,
),
),
(r"block_sparse_moe/mlp/(w1|w3)/kernel", pmag.resolve(ColumnWise)),
(r"block_sparse_moe/mlp/w2/kernel", pmag.resolve(RowWise)),
(r"residual_mlp/(w1|w3)/kernel", pmag.resolve(ColumnWise)),
(r"residual_mlp/w2/kernel", pmag.resolve(RowWise)),
(r"lm_head/kernel", pmag.resolve(ColumnWise)),
(r"score/kernel", pmag.resolve(RowWise)),
(r".*", pmag.resolve(Replicated)),
)
[docs] def get_mask_details(self) -> dict[int, AttnMaskDetail]:
"""Retrieve attention mask details for each layer in the model.
This method generates a dictionary mapping layer indices to their corresponding attention mask details.
If a sliding window is defined, each layer is assigned a sliding window attention mask with the specified size.
Returns:
dict[int, AttnMaskDetail]: A dictionary where keys are layer indices (int) and values are AttnMaskDetail
objects specifying the attention mask type and size for each layer.
Notes:
- If `self.sliding_window` is None, an empty dictionary is returned.
- The method iterates over `self.num_hidden_layers` to assign mask details for each layer.
- The attention mask type is set to `AttnMaskType.SLIDING` when a sliding window is defined.
"""
mapping = {}
if self.layer_types is not None:
for layer_idx in range(self.num_hidden_layers):
mapping[layer_idx] = AttnMaskDetail(
mask_type=AttnMaskType.from_hf(self.layer_types[layer_idx]),
size=self.sliding_window,
)
return mapping