Source code for easydel.modules.llama.llama_configuration

# 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.


from eformer.common_types import ColumnWise, Replicated, RowWise

from easydel.infra.base_module import EasyDeLBaseConfig
from easydel.infra.etils import EasyDeLGradientCheckPointers
from easydel.infra.factory import register_config


[docs]@register_config("llama") class LlamaConfig(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 Llama 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 11008): 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. number_rep_kv (`int`, *optional*, defaults to 1): Number of repetitions for the key and value vectors. num_key_value_heads (`int`, *optional*): Number of key and value heads for each attention layer in the Transformer encoder. Will default to `number_rep_kv * num_attention_heads` if not set. max_position_embeddings (`int`, *optional*, defaults to 2048): 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). head_dim (`int`, *optional*): head_dim for attention qkv. rms_norm_eps (`float`, *optional*, defaults to 1e-6): The epsilon used by the rms normalization layers. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. 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`. bos_token_id (`int`, *optional*, defaults to 0): The id of the *beginning-of-sequence* token. eos_token_id (`int`, *optional*, defaults to 1): The id of the *end-of-sequence* token. resid_pdrop (`float`, *optional*, defaults to 0.0): The dropout probability for all fully connected layers in the embeddings, encoder, and pooler. embd_pdrop (`float`, *optional*, defaults to 0.0): The dropout ratio for the embeddings. attention_dropout (`float`, *optional*, defaults to 0.0): The dropout ratio for the attention probabilities. rope_theta (`float`, *optional*, defaults to 10000.0): The theta value to use for rotary position embeddings. attention_bias (`bool`, *optional*, defaults to `False`): Whether to use attention bias. tie_word_embeddings (`bool`, *optional*, defaults to `False`): Whether to tie the weights of the input embeddings and the output embeddings. gradient_checkpointing (`str`, *optional*, defaults to `"nothing_saveable"`): The gradient checkpointing configuration. fcm_min_ratio (`float`, *optional*, defaults to -1): The minimum ratio for Flash Attention. fcm_max_ratio (`float`, *optional*, defaults to -1): The maximum ratio for Flash Attention. rope_scaling (`tp.Dict[str, tp.Union[str, float]]`, *optional*): The configuration for rope scaling. scan_mlp_chunk_size (`int`, *optional*, defaults to 1024): The chunk size to use when scanning the MLP. bits (`int`, *optional*): The number of bits to quantize the model to. hidden_act (`str`, *optional*, defaults to `"silu"`): The hidden activation function to use. pretraining_tp (`int`, *optional*, defaults to 1): The tensor parallelism degree used during pretraining. mlp_bias (`bool`, *optional*, defaults to `False`): Whether to use bias in the MLP. scan_layers (`bool`, *optional*, defaults to `False`): Whether to use the scan implementation for the layers. """ model_type: str = "llama" def __init__( self, vocab_size: int = 32000, hidden_size: int = 4096, intermediate_size: int = 11008, num_hidden_layers: int = 32, num_attention_heads: int = 32, number_rep_kv: int = 1, head_dim: int | None = None, num_key_value_heads: int | None = None, max_position_embeddings: int = 2048, rms_norm_eps: float = 1e-6, initializer_range: float = 0.02, use_cache: bool = True, bos_token_id: int = 0, eos_token_id: int = 1, resid_pdrop: float = 0.0, embd_pdrop: float = 0.0, attention_dropout: float = 0.0, rope_theta: float = 10000.0, attention_bias: bool = False, tie_word_embeddings: bool = False, gradient_checkpointing: EasyDeLGradientCheckPointers = EasyDeLGradientCheckPointers.NONE, fcm_min_ratio: float = -1, fcm_max_ratio: float = -1, rope_scaling: dict[str, str | float] | None = None, scan_mlp_chunk_size: int = 1024, bits: int | None = None, hidden_act: str = "silu", pretraining_tp: int = 1, mlp_bias: bool = False, scan_layers: bool = False, **kwargs, ): num_key_value_heads = num_key_value_heads or number_rep_kv * num_attention_heads self.num_key_value_heads = num_key_value_heads self.vocab_size = vocab_size self.number_rep_kv = number_rep_kv self.hidden_size = hidden_size self.initializer_range = initializer_range self.intermediate_size = intermediate_size self.num_hidden_layers = num_hidden_layers self.rope_theta = rope_theta self.attention_bias = attention_bias self.num_attention_heads = num_attention_heads self.max_position_embeddings = max_position_embeddings self.rms_norm_eps = rms_norm_eps self.use_cache = use_cache self.pretraining_tp = pretraining_tp self.resid_pdrop = resid_pdrop self.embd_pdrop = embd_pdrop self.attention_dropout = attention_dropout self.gradient_checkpointing = gradient_checkpointing self.mlp_bias = mlp_bias self.fcm_min_ratio = fcm_min_ratio self.hidden_act = hidden_act self.fcm_max_ratio = fcm_max_ratio self.rope_scaling = rope_scaling self.bits = bits self.scan_layers = scan_layers self.head_dim = head_dim if head_dim is not None else hidden_size // num_attention_heads super().__init__( bos_token_id=bos_token_id, eos_token_id=eos_token_id, tie_word_embeddings=tie_word_embeddings, scan_mlp_chunk_size=scan_mlp_chunk_size, bits=bits, **kwargs, )
[docs] def get_partition_rules(self, *args, **kwargs): """ Get the partition rules for the Llama model. Returns: `tp.Tuple[tp.Tuple[str, PartitionSpec]]`: The partition rules. """ pmag = self.partition_manager return ( (r"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"self_attn/.*proj/bias", pmag.resolve(Replicated)), (r"mlp/(gate_proj|up_proj)/kernel", pmag.resolve(ColumnWise)), (r"mlp/down_proj/kernel", pmag.resolve(RowWise)), (r"mlp/.*proj/bias", pmag.resolve(Replicated)), (r".*(input_layernorm|post_attention_layernorm|norm)/kernel", pmag.resolve(Replicated)), (r"lm_head/kernel", pmag.resolve(ColumnWise)), (r"score/kernel", pmag.resolve(RowWise)), (r".*bias", pmag.resolve(Replicated)), (r".*", pmag.resolve(Replicated)), )
[docs]class VisionLlamaConfig(LlamaConfig): """Configuration for Llama models augmented with a vision vocabulary.""" def __init__( self, vision_vocab_size=8448, tie_vision_embeddings=False, sample_mode="all", **kwargs, ): super().__init__(**kwargs) self.vision_vocab_size = vision_vocab_size self.tie_vision_embeddings = tie_vision_embeddings self.sample_mode = sample_mode
[docs] def get_partition_rules(self, *args, **kwargs): """ Get the partition rules for the Llama model. Returns: `tp.Tuple[tp.Tuple[str, PartitionSpec]]`: The partition rules. """ pmag = self.partition_manager return ( # Column-wise Sharding (split output dimensions) (r".*attn/.*(q_proj|k_proj|v_proj)/kernel", pmag.resolve(ColumnWise)), # QKV Projections (r".*mlp/(gate_proj|up_proj)/kernel", pmag.resolve(ColumnWise)), # MLP Up-Projections (r".*embed_tokens/embedding", pmag.resolve(ColumnWise)), # Token Embeddings (r".*embed_vision/embedding", pmag.resolve(ColumnWise)), # Vision Embeddings (r".*lm_head/kernel", pmag.resolve(ColumnWise)), # Language Model Head (r".*vision_head/kernel", pmag.resolve(ColumnWise)), # Vision Model Head # Row-wise Sharding (split input dimensions) (r".*attn/o_proj/kernel", pmag.resolve(RowWise)), # Attention Output (r".*mlp/down_proj/kernel", pmag.resolve(RowWise)), # MLP Down-Projection (r".*score/kernel", pmag.resolve(RowWise)), # Sequence Classifier Head # Replicated Parameters (r".*bias", pmag.resolve(Replicated)), # All biases (r".*layernorm/scale", pmag.resolve(Replicated)), # LayerNorm scales (r".*rms_norm/scale", pmag.resolve(Replicated)), # RMSNorm scales (r".*norm/scale", pmag.resolve(Replicated)), # Final LayerNorm scale (r".*", pmag.resolve(Replicated)), )