mlx-examples/llms/mlx_lm/models/phi.py

173 lines
5.4 KiB
Python
Raw Normal View History

# Copyright © 2023-2024 Apple Inc.
import math
from dataclasses import dataclass
from typing import Tuple
import mlx.core as mx
import mlx.nn as nn
Unify attention mask in LLMs (#911) * Unify attention mask creation in LLMs. Currently, each model implementation in `mlx-examples/llms/models` has ad-hoc code to create a mask for the attention mechanism. This usually takes the form: ``` mask = None if h.shape[1] > 1: mask = nn.MultiHeadAttention.create_additive_causal_mask(h.shape[1]) mask = mask.astype(h.dtype) ``` This correctly creates a mask only if the input consists of more than one token. But this code assumes the multi-token input is at the beginning of inference. If, for example, we are evaluating multiple tokens because of speculative decoding or prompt cache reuse, this mask will not have the correct shape and and will cause the raising of an exception in the attention computation. Some of the models correctly implement the mask creation with code like this: ``` mask = None if h.shape[1] > 1: mask = create_additive_causal_mask( h.shape[1], cache[0].offset if cache is not None else 0 ) mask = mask.astype(h.dtype) ``` This commit unifies the attention mask creation for all models with a new function `create_attention_mask`, reducing code duplication and helping all models support inference performance enhancements like those mentioned above. * Allow batches in LLM key-value cache The current implementation of the LLM key-value cache assumes that the input batch is of size 1. Input batching (evaluating multiple alterative inputs at the same time) can be a valuable tool for speculative sampling and other techniques. This change removes the hard-coded batch size from the code that resizes the key-value cache. * Simplify causal mask creation Use the same codepath regardless of whether there's an offset or not. Addresses [this comment](https://github.com/ml-explore/mlx-examples/pull/911#discussion_r1691459717). * Use old-style type annotation to avoid linter error
2024-07-26 07:45:22 +08:00
from .base import BaseModelArgs, create_attention_mask
@dataclass
class ModelArgs(BaseModelArgs):
model_type: str = "phi"
max_position_embeddings: int = 2048
vocab_size: int = 51200
hidden_size: int = 2560
num_attention_heads: int = 32
num_hidden_layers: int = 32
num_key_value_heads: int = 32
partial_rotary_factor: float = 0.4
intermediate_size: int = 10240
layer_norm_eps: float = 1e-5
rope_theta: float = 10000.0
def __post_init__(self):
if self.num_key_value_heads is None:
self.num_key_value_heads = self.num_attention_heads
class PhiAttention(nn.Module):
def __init__(self, config: ModelArgs):
super().__init__()
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.repeats = self.num_heads // self.num_key_value_heads
self.rope_theta = config.rope_theta
self.partial_rotary_factor = config.partial_rotary_factor
if (self.head_dim * self.num_heads) != self.hidden_size:
raise ValueError(
f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
f" and `num_heads`: {self.num_heads})."
)
self.q_proj = nn.Linear(
self.hidden_size, self.num_heads * self.head_dim, bias=True
)
self.k_proj = nn.Linear(
self.hidden_size, self.num_key_value_heads * self.head_dim, bias=True
)
self.v_proj = nn.Linear(
self.hidden_size, self.num_key_value_heads * self.head_dim, bias=True
)
self.dense = nn.Linear(
self.num_heads * self.head_dim, self.hidden_size, bias=True
)
self.rope = nn.RoPE(
int(self.partial_rotary_factor * self.head_dim),
traditional=False,
base=self.rope_theta,
)
def __call__(self, x, mask=None, cache=None):
queries, keys, values = self.q_proj(x), self.k_proj(x), self.v_proj(x)
# Extract some shapes
B, L, D = queries.shape
n_heads, n_kv_heads = self.num_heads, self.num_key_value_heads
# Prepare the queries, keys and values for the attention computation
queries = queries.reshape(
B,
L,
n_heads,
-1,
).moveaxis(1, 2)
keys = keys.reshape(B, L, n_kv_heads, -1).moveaxis(1, 2)
values = values.reshape(B, L, n_kv_heads, -1).moveaxis(1, 2)
# Add RoPE to the queries and keys and combine them with the cache
if cache is not None:
queries = self.rope(queries, offset=cache.offset)
keys = self.rope(keys, offset=cache.offset)
keys, values = cache.update_and_fetch(keys, values)
else:
queries = self.rope(queries)
keys = self.rope(keys)
scale = math.sqrt(1 / queries.shape[-1])
output = mx.fast.scaled_dot_product_attention(
queries.astype(mx.float32), keys, values, scale=scale, mask=mask
).astype(values.dtype)
output = output.moveaxis(2, 1).reshape(B, L, -1)
return self.dense(output)
class PhiMLP(nn.Module):
def __init__(self, config: ModelArgs):
super().__init__()
self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)
self.act = nn.GELU(approx="precise")
def __call__(self, x) -> mx.array:
return self.fc2(self.act(self.fc1(x)))
class PhiDecoderLayer(nn.Module):
def __init__(self, config: ModelArgs):
super().__init__()
self.self_attn = PhiAttention(config=config)
self.input_layernorm = nn.LayerNorm(
config.hidden_size, eps=config.layer_norm_eps
)
self.mlp = PhiMLP(config)
def __call__(self, x, mask, cache):
h = self.input_layernorm(x)
attn_h = self.self_attn(h, mask, cache)
ff_h = self.mlp(h)
return attn_h + ff_h + x
class PhiModel(nn.Module):
def __init__(self, config: ModelArgs):
super().__init__()
self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size)
self.layers = [PhiDecoderLayer(config) for i in range(config.num_hidden_layers)]
self.final_layernorm = nn.LayerNorm(
config.hidden_size, eps=config.layer_norm_eps
)
def __call__(self, x, cache):
x = self.embed_tokens(x)
Unify attention mask in LLMs (#911) * Unify attention mask creation in LLMs. Currently, each model implementation in `mlx-examples/llms/models` has ad-hoc code to create a mask for the attention mechanism. This usually takes the form: ``` mask = None if h.shape[1] > 1: mask = nn.MultiHeadAttention.create_additive_causal_mask(h.shape[1]) mask = mask.astype(h.dtype) ``` This correctly creates a mask only if the input consists of more than one token. But this code assumes the multi-token input is at the beginning of inference. If, for example, we are evaluating multiple tokens because of speculative decoding or prompt cache reuse, this mask will not have the correct shape and and will cause the raising of an exception in the attention computation. Some of the models correctly implement the mask creation with code like this: ``` mask = None if h.shape[1] > 1: mask = create_additive_causal_mask( h.shape[1], cache[0].offset if cache is not None else 0 ) mask = mask.astype(h.dtype) ``` This commit unifies the attention mask creation for all models with a new function `create_attention_mask`, reducing code duplication and helping all models support inference performance enhancements like those mentioned above. * Allow batches in LLM key-value cache The current implementation of the LLM key-value cache assumes that the input batch is of size 1. Input batching (evaluating multiple alterative inputs at the same time) can be a valuable tool for speculative sampling and other techniques. This change removes the hard-coded batch size from the code that resizes the key-value cache. * Simplify causal mask creation Use the same codepath regardless of whether there's an offset or not. Addresses [this comment](https://github.com/ml-explore/mlx-examples/pull/911#discussion_r1691459717). * Use old-style type annotation to avoid linter error
2024-07-26 07:45:22 +08:00
mask = create_attention_mask(x, cache)
if cache is None:
cache = [None] * len(self.layers)
for layer, c in zip(self.layers, cache):
x = layer(x, mask, c)
return self.final_layernorm(x)
class Model(nn.Module):
def __init__(self, config: ModelArgs):
super().__init__()
self.model_type = config.model_type
self.model = PhiModel(config)
self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=True)
self.args = config
def __call__(
self,
x: mx.array,
cache=None,
) -> mx.array:
y = self.model(x, cache)
return self.lm_head(y)
@property
def layers(self):
return self.model.layers