# Copyright 2023 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 functools import cached_property, partial
import chex
import jax
import jax.numpy as jnp
from eformer import common_types
from eformer.escale import apply_logical_sharding
from flax import nnx as nn
from easydel.infra.base_module import EasyDeLBaseModule
from easydel.infra.factory import TaskType, register_module
from easydel.infra.modeling_outputs import (
AttentionLayerOutput,
BaseModelOutput,
CausalLMOutput,
DecoderLayerOutput,
)
from easydel.infra.utils import (
ACT2FN,
auto_remat,
block_wise_ffn,
get_dot_general_by_bits,
)
from easydel.layers.attention import AttentionModule, FlexibleAttentionModule
from easydel.layers.caching import (
PagedAttentionCache,
PagedAttentionCacheView,
PagedAttentionMetadata,
TransformerCache,
TransformerCacheView,
TransformerMetadata,
)
from easydel.layers.linear import ParallelLinear
from .stablelm_configuration import StableLmConfig
[docs]class StableLmMLP(nn.Module):
"""Multi-Layer Perceptron (MLP) block for the StableLM model.
Attributes:
config (StableLmConfig): Configuration object for the model.
gate_proj (ParallelLinear): Linear layer for the gating mechanism.
down_proj (ParallelLinear): Linear layer for down-projection.
up_proj (ParallelLinear): Linear layer for up-projection.
act_fn (callable): Activation function (specified in config).
dtype (jnp.dtype): Data type for computations.
param_dtype (jnp.dtype): Data type for parameters.
precision (jax.lax.PrecisionLike): Precision setting for matrix multiplications.
"""
def __init__(
self,
config: StableLmConfig,
dtype: jnp.dtype = jnp.float32,
param_dtype: jnp.dtype = jnp.float32,
precision: jax.lax.PrecisionLike = None,
*,
rngs: nn.Rngs,
):
"""Initializes the StableLmMLP module.
Args:
config (StableLmConfig): The configuration object for the model.
dtype (jnp.dtype): Data type for computations (default: jnp.float32).
param_dtype (jnp.dtype): Data type for parameters (default: jnp.float32).
precision (jax.lax.PrecisionLike): Precision setting for JAX operations (default: None).
rngs (nn.Rngs): Random number generators.
"""
self.config = config
self.dtype = dtype
self.param_dtype = param_dtype
self.precision = precision
linear_class = partial(
ParallelLinear,
dtype=dtype,
param_dtype=param_dtype,
use_bias=False,
kernel_init=jax.nn.initializers.normal(config.initializer_range),
precision=precision,
**get_dot_general_by_bits(config.bits, config.easy_method),
)
self.gate_proj = linear_class(
config.hidden_size,
config.intermediate_size,
rngs=rngs,
)
self.down_proj = linear_class(
config.intermediate_size,
config.hidden_size,
rngs=rngs,
)
self.up_proj = linear_class(
config.hidden_size,
config.intermediate_size,
rngs=rngs,
)
self.act_fn = ACT2FN[config.hidden_act]
def __call__(self, hidden_states: jnp.ndarray) -> jnp.ndarray:
"""Forward pass of the MLP block.
Args:
hidden_states (jnp.ndarray): Input hidden states.
Returns:
jnp.ndarray: Output hidden states after MLP transformation.
"""
hidden_states = apply_logical_sharding(
hidden_states,
dynamic_axes=common_types.HiddenStateSharding,
partition_manager=self.config.partition_manager,
)
gate = self.act_fn(self.gate_proj(hidden_states))
up = self.up_proj(hidden_states)
hidden_states = self.down_proj(gate * up)
hidden_states = apply_logical_sharding(
hidden_states,
dynamic_axes=common_types.HiddenStateSharding,
partition_manager=self.config.partition_manager,
)
return hidden_states
[docs]class StableLmLayerNormPerHead(nn.Module):
"""Applies Layer Normalization independently to each attention head's dimension.
Attributes:
norms (list[nn.LayerNorm]): List of LayerNorm modules, one per head.
"""
def __init__(
self,
head_dim: int,
num_heads: int,
eps: float = 1e-5,
bias: bool = False,
dtype: jnp.dtype = jnp.float32,
param_dtype: jnp.dtype = jnp.float32,
*,
rngs: nn.Rngs,
):
"""Initializes the StableLmLayerNormPerHead module.
Args:
head_dim (int): The dimension of each attention head.
num_heads (int): The number of attention heads.
eps (float): Epsilon value for LayerNorm (default: 1e-5).
bias (bool): Whether to include bias in LayerNorm (default: False).
dtype (jnp.dtype): Data type for computations (default: jnp.float32).
param_dtype (jnp.dtype): Data type for parameters (default: jnp.float32).
rngs (nn.Rngs): Random number generators.
"""
self.norms = [
nn.LayerNorm(
head_dim,
epsilon=eps,
use_bias=bias,
dtype=dtype,
param_dtype=param_dtype,
rngs=rngs,
)
for idx in range(num_heads)
]
def __call__(self, hidden_states):
"""Applies LayerNorm per head.
Args:
hidden_states (chex.Array): Input hidden states, expected shape (..., num_heads * head_dim).
Returns:
chex.Array: Hidden states after applying LayerNorm per head, same shape as input.
"""
# hidden_states: [batch, seq_len, num_heads * head_dim]
# Reshape to [batch, seq_len, num_heads, head_dim]
states_per_heads = jnp.split(hidden_states, 1, axis=1)
# Normalize and merge the heads back together
return jnp.concatenate(
[
norm(hidden_states) for norm, hidden_states in zip(self.norms, states_per_heads)
],
axis=1,
)
[docs]class StableLmAttention(AttentionModule):
"""StableLM Attention module with Rotary Position Embeddings and optional LayerNorm on QK.
Attributes:
config (StableLmConfig): Configuration object for the model.
hidden_size (int): Dimensionality of the hidden states.
num_heads (int): Number of attention heads.
head_dim (int): Dimensionality of each attention head.
num_key_value_heads (int): Number of key/value heads (for GQA).
num_key_value_groups (int): Number of query heads per key/value head.
max_position_embeddings (int): Maximum sequence length.
rope_theta (float): Base value for RoPE.
partial_rotary_factor (float): Factor determining the portion of head dimension subject to RoPE.
q_proj (ParallelLinear): Linear layer for query projection.
k_proj (ParallelLinear): Linear layer for key projection.
v_proj (ParallelLinear): Linear layer for value projection.
o_proj (ParallelLinear): Linear layer for output projection.
rotary_emb_dim (int): Dimensionality of the rotary embeddings.
attention_performer (FlexibleAttentionModule): Module for performing attention computation.
qk_layernorm (bool): Whether to apply LayerNorm to query and key states.
q_layernorm (StableLmLayerNormPerHead): LayerNorm for query states (if qk_layernorm is True).
k_layernorm (StableLmLayerNormPerHead): LayerNorm for key states (if qk_layernorm is True).
rotary (RotaryEmbedding): Rotary positional embedding module.
dtype (jnp.dtype): Data type for computations.
param_dtype (jnp.dtype): Data type for parameters.
precision (jax.lax.PrecisionLike): Precision setting for matrix multiplications.
rngs (nn.Rngs): Random number generators.
"""
def __init__(
self,
config: StableLmConfig,
dtype: jnp.dtype = jnp.float32,
param_dtype: jnp.dtype = jnp.float32,
precision: jax.lax.PrecisionLike = None,
*,
rngs: nn.Rngs,
):
"""Initializes the StableLmAttention module.
Args:
config (StableLmConfig): The configuration object for the model.
dtype (jnp.dtype): Data type for computations (default: jnp.float32).
param_dtype (jnp.dtype): Data type for parameters (default: jnp.float32).
precision (jax.lax.PrecisionLike): Precision setting for JAX operations (default: None).
rngs (nn.Rngs): Random number generators.
"""
super().__init__(config=config)
self.dtype = dtype
self.param_dtype = param_dtype
self.precision = precision
self.rngs = rngs
self.hidden_size = config.hidden_size
self.num_heads = config.num_attention_heads
self.head_dim = self.hidden_size // self.num_heads
self.num_key_value_heads = config.num_key_value_heads
self.num_key_value_groups = self.num_heads // self.num_key_value_heads
self.max_position_embeddings = config.max_position_embeddings
self.rope_theta = config.rope_theta
self.partial_rotary_factor = config.partial_rotary_factor
if self.num_key_value_groups == 1:
assert self.config.num_attention_heads == self.config.num_key_value_heads
linear_class = partial(
ParallelLinear,
dtype=dtype,
param_dtype=param_dtype,
kernel_init=jax.nn.initializers.normal(config.initializer_range),
precision=precision,
**get_dot_general_by_bits(config.bits, config.easy_method),
)
self.q_proj = linear_class(
config.hidden_size,
config.num_attention_heads * self.head_dim,
use_bias=self.config.use_qkv_bias,
rngs=rngs,
)
self.k_proj = linear_class(
config.hidden_size,
config.num_key_value_heads * self.head_dim,
use_bias=self.config.use_qkv_bias,
rngs=rngs,
)
self.v_proj = linear_class(
config.hidden_size,
config.num_key_value_heads * self.head_dim,
use_bias=self.config.use_qkv_bias,
rngs=rngs,
)
self.o_proj = linear_class(
config.num_attention_heads * self.head_dim,
config.hidden_size,
use_bias=False,
rngs=rngs,
)
self.rotary_emb_dim = int(self.config.partial_rotary_factor * self.head_dim)
self.attention_performer = FlexibleAttentionModule(
base_config=config,
softmax_scale=self.head_dim**-0.5,
dropout_prob=config.attention_dropout,
)
self.qk_layernorm = config.qk_layernorm
if self.qk_layernorm:
self.q_layernorm = StableLmLayerNormPerHead(
head_dim=self.head_dim,
num_heads=config.num_attention_heads,
eps=config.layer_norm_eps,
dtype=self.dtype,
param_dtype=self.param_dtype,
rngs=rngs,
)
self.k_layernorm = StableLmLayerNormPerHead(
head_dim=self.head_dim,
num_heads=config.num_key_value_heads,
eps=config.layer_norm_eps,
dtype=self.dtype,
param_dtype=self.param_dtype,
rngs=rngs,
)
self.rotary = self.config.get_basic_rope(
self.dtype,
head_size=int(
config.partial_rotary_factor
* (config.hidden_size // config.num_attention_heads)
),
rotary_dim=self.rotary_emb_dim,
base=config.rope_theta,
)
def __call__(
self,
hidden_states: chex.Array,
attention_mask: chex.Array,
position_ids: chex.Array,
causal_mask: tp.Optional[chex.Array | bool],
mode: common_types.RUNTIME_MODE_TYPES, # type:ignore
cache_view: tp.Optional[TransformerCacheView | PagedAttentionCacheView] = None,
cache_metadata: tp.Optional[TransformerMetadata | PagedAttentionMetadata] = None,
segment_ids: tp.Optional[chex.Array] = None,
output_attentions: bool = False,
fcm_mask: tp.Optional[chex.Array] = None,
frequencies: tp.Optional[chex.Array] = None,
):
"""Forward pass of the attention module.
Args:
hidden_states (chex.Array): Input hidden states (batch, seq_len, hidden_size).
attention_mask (chex.Array): Mask to apply on the attention scores (batch, 1, seq_len, kv_seq_len).
position_ids (chex.Array): Position indices for the tokens (batch, seq_len).
causal_mask (tp.Optional[chex.Array | bool]): Causal mask for ensuring autoregressive behavior.
cache_view (tp.Optional[TransformerCacheView | PagedAttentionCacheView]): Cache view for key/value states (optional).
cache_metadata (tp.Optional[TransformerMetadata | PagedAttentionMetadata]): Metadata for paged attention (optional).
segment_ids (tp.Optional[chex.Array]): Segment IDs for segment-based attention (optional).
output_attentions (bool): If True, outputs attention weights alongside the hidden states (default: False).
fcm_mask (tp.Optional[chex.Array]): Forward causal mask (FCM) mask (optional).
frequencies (tp.Optional[chex.Array]): Precomputed rotary frequencies (optional).
Returns:
tp.Tuple[chex.Array, chex.Array | None]: A tuple containing the attention output (batch, seq_len, hidden_size)
and optionally the attention weights (batch, num_heads, seq_len, kv_seq_len).
"""
batch_size, sequence_length = hidden_states.shape[:2]
query_states, key_states, value_states = (
self.q_proj(hidden_states),
self.k_proj(hidden_states),
self.v_proj(hidden_states),
)
query_states = query_states.reshape(
batch_size,
sequence_length,
self.config.num_attention_heads,
self.head_dim,
)
key_states = key_states.reshape(
batch_size,
sequence_length,
self.config.num_key_value_heads,
self.head_dim,
)
value_states = value_states.reshape(
batch_size,
sequence_length,
self.config.num_key_value_heads,
self.head_dim,
)
if self.qk_layernorm:
query_states = self.q_layernorm(query_states.transpose(0, 2, 1, 3)).transpose(
0, 2, 1, 3
)
key_states = self.k_layernorm(key_states.transpose(0, 2, 1, 3)).transpose(
0, 2, 1, 3
)
(
query_states,
key_states,
value_states,
) = self.apply_qkv_shardings(query_states, key_states, value_states)
query_states, key_states = self.rotary(
positions=position_ids,
query=query_states,
key=key_states,
frequencies=frequencies,
)
(
key_states,
value_states,
attention_mask,
init_attention_bias,
cache_view,
) = self.concatenate(
query=query_states,
key=key_states,
value=value_states,
cache_view=cache_view,
cache_metadata=cache_metadata,
attention_mask=attention_mask,
causal_mask=causal_mask,
fcm_mask=fcm_mask,
)
attentions = self.attention_performer.forward(
query_states=query_states,
key_states=key_states,
value_states=value_states,
mode=mode,
bias=None,
cache_metadata=cache_metadata,
cache_view=cache_view,
init_bias=init_attention_bias,
attention_mask=attention_mask,
segment_ids=segment_ids,
causal=True,
dropout_rng=self.rngs.params(),
)
attn_output = self.shard_attention_prod(
self._merge_heads(attentions.attention_outputs)
)
attn_output = self.o_proj(attn_output)
return AttentionLayerOutput(
attention_output=attn_output,
attention_weight=attentions.attention_weights if output_attentions else None,
cache_view=cache_view,
)
[docs]class StableLmDecoderLayer(nn.Module):
"""A single decoder layer for the StableLM model.
This layer combines self-attention, MLP, and residual connections with layer normalization.
It supports parallel residual connections.
Attributes:
config (StableLmConfig): Configuration object for the model.
self_attn (StableLmAttention): Self-attention module.
mlp (StableLmMLP): MLP module.
input_layernorm (nn.LayerNorm): Layer normalization applied before self-attention.
post_attention_layernorm (nn.LayerNorm): Layer normalization applied after self-attention and before the MLP.
dropout_rng_key (str): Name of the RNG key for dropout.
dtype (jnp.dtype): Data type for computations.
param_dtype (jnp.dtype): Data type for parameters.
precision (jax.lax.PrecisionLike): Precision setting for matrix multiplications.
rngs (nn.Rngs): Random number generators.
"""
def __init__(
self,
config: StableLmConfig,
dtype: jnp.dtype = jnp.float32,
param_dtype: jnp.dtype = jnp.float32,
precision: jax.lax.PrecisionLike = None,
*,
rngs: nn.Rngs,
):
"""Initializes the StableLmDecoderLayer module.
Args:
config (StableLmConfig): The configuration object for the model.
dtype (jnp.dtype): Data type for computations (default: jnp.float32).
param_dtype (jnp.dtype): Data type for parameters (default: jnp.float32).
precision (jax.lax.PrecisionLike): Precision setting for JAX operations (default: None).
rngs (nn.Rngs): Random number generators.
"""
self.config = config
self.dtype = dtype
self.param_dtype = param_dtype
self.precision = precision
attn_block = StableLmAttention
mlp_block = StableLmMLP
self.use_parallel_residual = self.config.use_parallel_residual
attn_block, mlp_block = auto_remat(
attn_block,
mlp_block,
policy=config.gradient_checkpointing,
)
self.self_attn = attn_block(
config=config,
dtype=dtype,
param_dtype=param_dtype,
precision=precision,
rngs=rngs,
)
self.mlp = mlp_block(
config=config,
dtype=dtype,
param_dtype=param_dtype,
precision=precision,
rngs=rngs,
)
self.input_layernorm = nn.LayerNorm(
config.hidden_size,
epsilon=config.layer_norm_eps,
dtype=dtype,
param_dtype=param_dtype,
rngs=rngs,
)
if not self.use_parallel_residual:
self.post_attention_layernorm = nn.LayerNorm(
config.hidden_size,
epsilon=config.layer_norm_eps,
dtype=dtype,
param_dtype=param_dtype,
rngs=rngs,
)
self.dropout = nn.Dropout(self.config.hidden_dropout, rngs=rngs)
def __call__(
self,
hidden_states: chex.Array,
attention_mask: chex.Array,
position_ids: chex.Array,
causal_mask: tp.Optional[chex.Array | bool],
mode: common_types.RUNTIME_MODE_TYPES, # type:ignore
cache_view: tp.Optional[TransformerCacheView | PagedAttentionCacheView] = None,
cache_metadata: tp.Optional[TransformerMetadata | PagedAttentionMetadata] = None,
segment_ids: tp.Optional[chex.Array] = None,
output_attentions: bool = False,
fcm_mask: tp.Optional[chex.Array] = None,
frequencies: tp.Optional[chex.Array] = None,
):
"""Forward pass of the decoder layer.
Args:
hidden_states (chex.Array): Input hidden states (batch, seq_len, hidden_size).
attention_mask (chex.Array): Attention mask (batch, 1, seq_len, kv_seq_len).
position_ids (chex.Array): Position IDs (batch, seq_len).
causal_mask (tp.Optional[chex.Array | bool]): Causal mask for autoregressive behavior.
cache_view (tp.Optional[TransformerCacheView | PagedAttentionCacheView]): Cache view for key/value states (optional).
cache_metadata (tp.Optional[TransformerMetadata | PagedAttentionMetadata]): Metadata for paged attention (optional).
segment_ids (tp.Optional[chex.Array]): Segment IDs for segment-based attention (optional).
output_attentions (bool): Whether to output attention weights (default: False).
fcm_mask (tp.Optional[chex.Array]): Forward causal mask (FCM) mask (optional).
frequencies (tp.Optional[chex.Array]): Precomputed rotary frequencies (optional).
Returns:
tp.Tuple[chex.Array, chex.Array | None]: A tuple containing:
- hidden_states (chex.Array): Output hidden states after the decoder layer.
- attention_outputs (chex.Array | None): Attention weights (if `output_attentions` is True).
"""
assert hidden_states.ndim == 3, (
f"Input hidden_states should be 3 dimensions, got {hidden_states.ndim}"
)
residual = hidden_states
hidden_states = self.input_layernorm(hidden_states)
hidden_states = apply_logical_sharding(
hidden_states,
dynamic_axes=common_types.HiddenStateSharding,
partition_manager=self.config.partition_manager,
)
attn_outputs = self.self_attn(
hidden_states,
attention_mask,
position_ids,
causal_mask,
mode,
cache_view,
cache_metadata,
segment_ids,
output_attentions,
fcm_mask,
frequencies,
)
if self.use_parallel_residual:
if self.config.use_scan_mlp:
hidden_states = block_wise_ffn(
self.mlp, hidden_states, self.config.scan_mlp_chunk_size
)
else:
hidden_states = self.mlp(hidden_states)
hidden_states = self.dropout(hidden_states)
hidden_states = hidden_states + residual + attn_outputs.attention_output
else:
residual = residual + attn_outputs.attention_output
if self.config.use_scan_mlp:
hidden_states = block_wise_ffn(
self.mlp,
self.post_attention_layernorm(residual),
self.config.scan_mlp_chunk_size,
)
else:
hidden_states = self.mlp(self.post_attention_layernorm(residual))
hidden_states = self.dropout(hidden_states)
hidden_states = hidden_states + residual
hidden_states = apply_logical_sharding(
hidden_states,
dynamic_axes=common_types.HiddenStateSharding,
partition_manager=self.config.partition_manager,
)
return DecoderLayerOutput(
hidden_states=hidden_states,
attention_weight=attn_outputs.attention_weight,
cache_view=attn_outputs.cache_view,
)
[docs]@register_module(
TaskType.BASE_MODULE,
config=StableLmConfig,
model_type="stablelm",
)
class StableLmModel(EasyDeLBaseModule):
"""The base StableLM transformer model.
This class implements the core transformer architecture, including embedding layers,
decoder layers, and final normalization.
Attributes:
config (StableLmConfig): Configuration object for the model.
embed_tokens (nn.Embed): Embedding layer for input tokens.
layers (nn.List[StableLmDecoderLayer]): List of decoder layers.
norm (nn.LayerNorm): Final layer normalization.
gradient_checkpointing (str): Gradient checkpointing strategy.
dtype (jnp.dtype): Data type for computations.
param_dtype (jnp.dtype): Data type for parameters.
precision (jax.lax.PrecisionLike): Precision setting for matrix multiplications.
rngs (nn.Rngs): Random number generators.
"""
def __init__(
self,
config: StableLmConfig,
dtype: jnp.dtype = jnp.float32,
param_dtype: jnp.dtype = jnp.float32,
precision: jax.lax.PrecisionLike = None,
*,
rngs: nn.Rngs,
):
"""Initializes the StableLmModel module.
Args:
config (StableLmConfig): The configuration object for the model.
dtype (jnp.dtype): Data type for computations (default: jnp.float32).
param_dtype (jnp.dtype): Data type for parameters (default: jnp.float32).
precision (jax.lax.PrecisionLike): Precision setting for JAX operations (default: None).
rngs (nn.Rngs): Random number generators.
"""
super().__init__(
config=config,
dtype=dtype,
param_dtype=param_dtype,
precision=precision,
rngs=rngs,
)
self.padding_idx = config.pad_token_id
self.vocab_size = config.vocab_size
self.embed_tokens = nn.Embed(
config.vocab_size,
config.hidden_size,
dtype=dtype,
param_dtype=param_dtype,
rngs=rngs,
)
self.layers = [
StableLmDecoderLayer(
config=config,
dtype=dtype,
param_dtype=param_dtype,
precision=precision,
rngs=rngs,
)
for idx in range(config.num_hidden_layers)
]
self.norm = nn.LayerNorm(
config.hidden_size,
epsilon=config.layer_norm_eps,
dtype=dtype,
param_dtype=param_dtype,
rngs=rngs,
)
@cached_property
def frequencies(self):
"""Cached property for precomputed rotary frequencies."""
rotary_emb_dim = int(
self.config.partial_rotary_factor
* (self.config.hidden_size // self.config.num_attention_heads)
)
self._frequencies = self.config.get_basic_frequencies(
head_size=rotary_emb_dim,
rotary_dim=rotary_emb_dim,
)
def __call__(
self,
input_ids: tp.Optional[chex.Array] = None,
inputs_embeds: tp.Optional[chex.Array] = None,
attention_mask: tp.Optional[chex.Array] = None,
position_ids: tp.Optional[chex.Array] = None,
segment_ids: tp.Optional[chex.Array] = None,
output_attentions: tp.Optional[bool] = None,
output_hidden_states: tp.Optional[bool] = None,
mode: tp.Optional[common_types.RUNTIME_MODE_TYPES] = None, # type:ignore
past_key_values: tp.Optional[TransformerCache | PagedAttentionCache] = None,
cache_metadata: tp.Optional[TransformerMetadata | PagedAttentionMetadata] = None,
) -> BaseModelOutput:
"""Forward pass of the StableLM model.
Args:
input_ids (tp.Optional[chex.Array]): Input token IDs (batch, seq_len). Mutually exclusive with `inputs_embeds`.
inputs_embeds (tp.Optional[chex.Array]): Input embeddings (batch, seq_len, hidden_size). Mutually exclusive with `input_ids`.
attention_mask (tp.Optional[chex.Array]): Attention mask (batch, seq_len). Usually used for padding tokens.
position_ids (tp.Optional[chex.Array]): Position IDs (batch, seq_len). If None, automatically generated.
segment_ids (tp.Optional[chex.Array]): Segment IDs for segment-based attention (optional).
output_attentions (tp.Optional[bool]): Whether to output attention weights (default defined by config).
output_hidden_states (tp.Optional[bool]): Whether to output hidden states for all layers (default defined by config).
past_key_values (tp.Optional[TransformerCache | PagedAttentionCache]): Precomputed key/value states for caching.
cache_metadata (tp.Optional[TransformerMetadata | PagedAttentionMetadata]): Metadata for paged attention (optional).
Returns:
BaseModelOutput: The model output, either as a `BaseModelOutput` object or a tuple.
Raises:
ValueError: If both `input_ids` and `inputs_embeds` are provided or neither is provided.
"""
if (input_ids is None) ^ (inputs_embeds is not None):
raise ValueError(
"You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one"
)
if inputs_embeds is None:
inputs_embeds = self.embed_tokens(input_ids.astype("i4"))
batch_size, sequence_length, _ = inputs_embeds.shape
all_attentions = () if output_attentions else None
all_hidden_states = () if output_hidden_states else None
assert sequence_length <= self.config.max_position_embeddings, (
f"Maximum Position Embedding Reached ! (Excepted <= {self.config.max_position_embeddings} got {sequence_length})"
)
if attention_mask is None:
attention_mask = jnp.ones((batch_size, sequence_length), "b1")
else:
if attention_mask.dtype != jnp.bool:
attention_mask = jnp.astype(attention_mask == 1, "b1")
if position_ids is None:
position_ids = jnp.broadcast_to(
jnp.clip(jnp.cumsum(attention_mask, axis=-1) - 1, a_min=0),
(batch_size, sequence_length),
).astype(jnp.int32)
if attention_mask.ndim == 2:
attention_mask = jnp.expand_dims(attention_mask, (1, 2))
hidden_states = inputs_embeds
if mode is None:
mode = (
common_types.MODE_DECODE
if sequence_length == 1 and past_key_values is not None
else common_types.MODE_TRAIN
)
if past_key_values is None:
past_key_values = TransformerCache.init_empty(len(self.layers))
hidden_states = apply_logical_sharding(
hidden_states,
dynamic_axes=common_types.HiddenStateSharding,
partition_manager=self.config.partition_manager,
)
for idx, block in enumerate(self.layers):
if output_hidden_states:
all_hidden_states += (hidden_states,)
layer_outputs = block(
hidden_states=hidden_states,
attention_mask=attention_mask,
position_ids=position_ids,
mode=mode,
cache_view=past_key_values.views[idx],
cache_metadata=cache_metadata,
causal_mask=self.causal_mask,
output_attentions=output_attentions,
segment_ids=segment_ids,
frequencies=self.frequencies,
)
hidden_states = layer_outputs.hidden_states
if output_attentions:
all_attentions += (layer_outputs.attention_weight,)
past_key_values[idx] = layer_outputs.cache_view
hidden_states = self.norm(hidden_states)
if output_hidden_states:
all_hidden_states += (hidden_states,)
return BaseModelOutput(
last_hidden_state=hidden_states,
hidden_states=all_hidden_states,
attentions=all_attentions,
past_key_values=past_key_values,
)
[docs]@register_module(
TaskType.CAUSAL_LM,
config=StableLmConfig,
model_type="stablelm",
)
class StableLmForCausalLM(EasyDeLBaseModule):
"""StableLM model with a Causal Language Modeling (CLM) head.
This class wraps the base `StableLmModel` and adds a linear layer (language model head)
to predict the next token logits.
Attributes:
config (StableLmConfig): Configuration object for the model.
model (StableLmModel): The base StableLM model.
lm_head (ParallelLinear): The language model head (linear layer).
dtype (jnp.dtype): Data type for computations.
param_dtype (jnp.dtype): Data type for parameters.
precision (jax.lax.PrecisionLike): Precision setting for matrix multiplications.
rngs (nn.Rngs): Random number generators.
"""
def __init__(
self,
config: StableLmConfig,
dtype: jnp.dtype = jnp.float32,
param_dtype: jnp.dtype = jnp.float32,
precision: jax.lax.PrecisionLike = None,
*,
rngs: nn.Rngs,
):
"""Initializes the StableLmForCausalLM module.
Args:
config (StableLmConfig): The configuration object for the model.
dtype (jnp.dtype): Data type for computations (default: jnp.float32).
param_dtype (jnp.dtype): Data type for parameters (default: jnp.float32).
precision (jax.lax.PrecisionLike): Precision setting for JAX operations (default: None).
rngs (nn.Rngs): Random number generators.
"""
super().__init__(
config=config,
dtype=dtype,
param_dtype=param_dtype,
precision=precision,
rngs=rngs,
)
self.model = StableLmModel(
config=config,
dtype=dtype,
param_dtype=param_dtype,
precision=precision,
rngs=rngs,
)
self.vocab_size = self.config.vocab_size
self.lm_head = ParallelLinear(
config.hidden_size,
config.vocab_size,
use_bias=False,
kernel_init=jax.nn.initializers.normal(config.initializer_range),
dtype=dtype,
param_dtype=param_dtype,
precision=precision,
rngs=rngs,
)
def __call__(
self,
input_ids: tp.Optional[chex.Array] = None,
inputs_embeds: tp.Optional[chex.Array] = None,
attention_mask: tp.Optional[chex.Array] = None,
position_ids: tp.Optional[chex.Array] = None,
segment_ids: tp.Optional[chex.Array] = None,
output_attentions: tp.Optional[bool] = None,
output_hidden_states: tp.Optional[bool] = None,
mode: tp.Optional[common_types.RUNTIME_MODE_TYPES] = None, # type:ignore
past_key_values: tp.Optional[TransformerCache | PagedAttentionCache] = None,
cache_metadata: tp.Optional[TransformerMetadata | PagedAttentionMetadata] = None,
) -> CausalLMOutput:
"""Forward pass of the StableLM model for Causal Language Modeling.
Args:
input_ids (tp.Optional[chex.Array]): Input token IDs (batch, seq_len). Mutually exclusive with `inputs_embeds`.
inputs_embeds (tp.Optional[chex.Array]): Input embeddings (batch, seq_len, hidden_size). Mutually exclusive with `input_ids`.
attention_mask (tp.Optional[chex.Array]): Attention mask (batch, seq_len). Usually used for padding tokens.
position_ids (tp.Optional[chex.Array]): Position IDs (batch, seq_len). If None, automatically generated.
segment_ids (tp.Optional[chex.Array]): Segment IDs for segment-based attention (optional).
output_attentions (tp.Optional[bool]): Whether to output attention weights (default defined by config).
output_hidden_states (tp.Optional[bool]): Whether to output hidden states for all layers (default defined by config).
past_key_values (tp.Optional[TransformerCache | PagedAttentionCache]): Precomputed key/value states for caching.
cache_metadata (tp.Optional[TransformerMetadata | PagedAttentionMetadata]): Metadata for paged attention (optional).
Returns:
CausalLMOutput: The model output, including logits, hidden states, and attentions.
"""
outputs = self.model(
input_ids=input_ids,
attention_mask=attention_mask,
position_ids=position_ids,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
mode=mode,
past_key_values=past_key_values,
cache_metadata=cache_metadata,
inputs_embeds=inputs_embeds,
segment_ids=segment_ids,
)
hidden_states = outputs.last_hidden_state
hidden_states = apply_logical_sharding(
hidden_states,
dynamic_axes=common_types.HiddenStateSharding,
partition_manager=self.config.partition_manager,
)
if self.config.tie_word_embeddings:
lm_logits = jax.lax.dot_general(
hidden_states,
self.model.embed_tokens.embedding.value.T,
(((hidden_states.ndim - 1), (0,)), ((), ())),
)
else:
lm_logits = self.lm_head(hidden_states)
return CausalLMOutput(
logits=lm_logits,
hidden_states=outputs.hidden_states,
attentions=outputs.attentions,
past_key_values=outputs.past_key_values,
)