Rain-Bus
ffd2defdfc
Annotated 16 source files covering the full architecture: engine (scheduler, block manager, model runner), layers (attention, linear, sampler, etc.), model (qwen3), and utils. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
85 lines
2.7 KiB
Python
85 lines
2.7 KiB
Python
from functools import lru_cache
|
|
import torch
|
|
from torch import nn
|
|
|
|
|
|
def apply_rotary_emb(
|
|
x: torch.Tensor,
|
|
cos: torch.Tensor,
|
|
sin: torch.Tensor,
|
|
) -> torch.Tensor:
|
|
"""应用旋转位置编码（RoPE）。
|
|
|
|
将向量沿最后一维分成两半 (x1, x2)，然后做旋转变换:
|
|
y1 = x1 * cos - x2 * sin
|
|
y2 = x2 * cos + x1 * sin
|
|
这等价于在二维平面上将每个相邻的 (x1, x2) 对旋转 theta 角度，
|
|
其中 theta = position / (base^(2i/d))。
|
|
"""
|
|
x1, x2 = torch.chunk(x.float(), 2, dim=-1)
|
|
y1 = x1 * cos - x2 * sin
|
|
y2 = x2 * cos + x1 * sin
|
|
return torch.cat((y1, y2), dim=-1).to(x.dtype)
|
|
|
|
|
|
class RotaryEmbedding(nn.Module):
|
|
"""旋转位置编码（Rotary Position Embedding, RoPE）。
|
|
|
|
RoPE 通过旋转矩阵编码位置信息，使得注意力计算中
|
|
内积只依赖相对位置（q_i · k_j 只与 i-j 有关），
|
|
从而天然支持外推到更长序列。
|
|
|
|
预计算所有位置的 cos 和 sin 值并缓存，避免重复计算。
|
|
缓存形状: [max_position_embeddings, 1, head_dim]，中间维是 num_heads 的广播维度。
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
head_size: int,
|
|
rotary_dim: int,
|
|
max_position_embeddings: int,
|
|
base: float,
|
|
) -> None:
|
|
super().__init__()
|
|
self.head_size = head_size
|
|
assert rotary_dim == head_size
|
|
# 计算逆频率: 1 / (base^(2i/d)), i = 0, 1, ..., d/2-1
|
|
inv_freq = 1.0 / (base**(torch.arange(0, rotary_dim, 2, dtype=torch.float) / rotary_dim))
|
|
# 计算所有位置的频率: pos * inv_freq
|
|
t = torch.arange(max_position_embeddings, dtype=torch.float)
|
|
freqs = torch.einsum("i,j -> ij", t, inv_freq)
|
|
cos = freqs.cos()
|
|
sin = freqs.sin()
|
|
# 拼接 cos 和 sin，形状 [max_pos, 1, head_dim]
|
|
cache = torch.cat((cos, sin), dim=-1).unsqueeze_(1)
|
|
self.register_buffer("cos_sin_cache", cache, persistent=False)
|
|
|
|
@torch.compile
|
|
def forward(
|
|
self,
|
|
positions: torch.Tensor,
|
|
query: torch.Tensor,
|
|
key: torch.Tensor,
|
|
) -> tuple[torch.Tensor, torch.Tensor]:
|
|
"""根据位置索引查找 cos/sin 并应用到 Q 和 K。"""
|
|
cos_sin = self.cos_sin_cache[positions]
|
|
cos, sin = cos_sin.chunk(2, dim=-1)
|
|
query = apply_rotary_emb(query, cos, sin)
|
|
key = apply_rotary_emb(key, cos, sin)
|
|
return query, key
|
|
|
|
|
|
@lru_cache(1)
|
|
def get_rope(
|
|
head_size: int,
|
|
rotary_dim: int,
|
|
max_position: int,
|
|
base: float,
|
|
):
|
|
"""获取 RotaryEmbedding 的单例实例。
|
|
|
|
使用 lru_cache 确保相同参数只创建一次。
|
|
"""
|
|
rotary_emb = RotaryEmbedding(head_size, rotary_dim, max_position, base)
|
|
return rotary_emb
|