File size: 10,633 Bytes
07d61c1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 |
# coding=utf-8
# Copyright 2025-present, the HuggingFace Inc. Team and AIRAS Inc. Team. All rights reserved.
#
# 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
#
# http://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 torch
import torch.nn as nn
import torch.nn.functional as F
from typing import Optional, Tuple
def precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0) -> torch.Tensor:
"""Precompute the frequency tensor for complex rotation."""
freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))
t = torch.arange(end, device=freqs.device)
freqs = torch.outer(t, freqs)
return torch.polar(torch.ones_like(freqs), freqs)
def apply_rotary_emb(x: torch.Tensor, freqs_cis: torch.Tensor) -> torch.Tensor:
"""Apply rotary position embeddings to the input tensor."""
x_complex = torch.view_as_complex(x.float().reshape(*x.shape[:-1], -1, 2))
freqs_cis = freqs_cis.view(1, *freqs_cis.shape)
x_rotated = x_complex * freqs_cis
return torch.view_as_real(x_rotated).flatten(-2)
class SapnousAttention(nn.Module):
"""Multi-head attention with rotary position embeddings and sliding window attention."""
def __init__(self, config):
super().__init__()
self.config = config
self.hidden_size = config.hidden_size
self.num_attention_heads = config.num_attention_heads
self.head_dim = self.hidden_size // self.num_attention_heads
self.num_key_value_heads = config.num_key_value_heads
self.num_key_value_groups = self.num_attention_heads // self.num_key_value_heads
self.max_position_embeddings = config.max_position_embeddings
self.rope_theta = config.rope_theta
self.sliding_window = config.sliding_window if config.use_sliding_window else None
if (self.head_dim * self.num_attention_heads) != self.hidden_size:
raise ValueError(
f"hidden_size must be divisible by num_attention_heads (got {self.hidden_size} and {self.num_attention_heads})"
)
self.q_proj = nn.Linear(self.hidden_size, self.num_attention_heads * self.head_dim, bias=False)
self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
self.o_proj = nn.Linear(self.num_attention_heads * self.head_dim, self.hidden_size, bias=False)
self.attention_dropout = nn.Dropout(config.attention_dropout)
def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int) -> torch.Tensor:
return tensor.view(bsz, seq_len, self.num_attention_heads, self.head_dim).transpose(1, 2)
def _kv_shape(self, tensor: torch.Tensor, seq_len: int, bsz: int) -> torch.Tensor:
return tensor.view(bsz, seq_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
def forward(
self,
hidden_states: torch.Tensor,
freqs_cis: torch.Tensor,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_value: Optional[Tuple[torch.Tensor]] = None,
output_attentions: bool = False,
use_cache: bool = False,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
bsz, q_len, _ = hidden_states.size()
query_states = self.q_proj(hidden_states)
key_states = self.k_proj(hidden_states)
value_states = self.v_proj(hidden_states)
query_states = self._shape(query_states, q_len, bsz)
key_states = self._kv_shape(key_states, q_len, bsz)
value_states = self._kv_shape(value_states, q_len, bsz)
kv_seq_len = key_states.shape[-2]
if past_key_value is not None:
kv_seq_len += past_key_value[0].shape[-2]
# Apply rotary position embeddings
if position_ids is None:
position_ids = torch.arange(kv_seq_len, device=hidden_states.device)
cos, sin = freqs_cis[position_ids]
query_states, key_states = apply_rotary_emb(query_states, cos), apply_rotary_emb(key_states, sin)
if past_key_value is not None:
# Reuse k, v, self_attention
key_states = torch.cat([past_key_value[0], key_states], dim=2)
value_states = torch.cat([past_key_value[1], value_states], dim=2)
past_key_value = (key_states, value_states) if use_cache else None
# Repeat k/v heads if n_kv_heads < n_heads
key_states = torch.repeat_interleave(key_states, self.num_key_value_groups, dim=1)
value_states = torch.repeat_interleave(value_states, self.num_key_value_groups, dim=1)
attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
if attention_mask is not None:
attn_weights = attn_weights + attention_mask
# Sliding window attention if configured
if self.sliding_window is not None and kv_seq_len > self.sliding_window:
# Create sliding window mask
window_mask = torch.ones_like(attn_weights, dtype=torch.bool)
for i in range(q_len):
window_start = max(0, i - self.sliding_window // 2)
window_end = min(kv_seq_len, i + self.sliding_window // 2)
window_mask[:, :, i, window_start:window_end] = False
attn_weights = attn_weights.masked_fill(window_mask, float('-inf'))
# Causal mask for autoregressive generation
if self.config.scoring_func == "softmax":
causal_mask = torch.triu(torch.ones((q_len, kv_seq_len), dtype=torch.bool), diagonal=1)
causal_mask = causal_mask.unsqueeze(0).unsqueeze(0)
attn_weights = attn_weights.masked_fill(causal_mask.to(attn_weights.device), float('-inf'))
attn_weights = F.softmax(attn_weights, dim=-1)
else:
# Alternative scoring functions (e.g., RoPE-only, cosine similarity)
attn_weights = F.relu(attn_weights)
attn_weights = attn_weights / (attn_weights.sum(dim=-1, keepdim=True) + 1e-6)
attn_weights = self.attention_dropout(attn_weights)
attn_output = torch.matmul(attn_weights, value_states)
attn_output = attn_output.transpose(1, 2).contiguous()
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
attn_output = self.o_proj(attn_output)
if not output_attentions:
attn_weights = None
return attn_output, attn_weights, past_key_value
class SapnousBlock(nn.Module):
"""Transformer block with attention, layer norm, and feed-forward network."""
def __init__(self, config):
super().__init__()
self.hidden_size = config.hidden_size
self.self_attn = SapnousAttention(config)
self.input_layernorm = nn.LayerNorm(config.hidden_size, eps=config.rms_norm_eps)
self.post_attention_layernorm = nn.LayerNorm(config.hidden_size, eps=config.rms_norm_eps)
self.mlp = nn.Sequential(
nn.Linear(config.hidden_size, config.intermediate_size, bias=False),
nn.SiLU(),
nn.Linear(config.intermediate_size, config.hidden_size, bias=False),
)
def forward(
self,
hidden_states: torch.Tensor,
freqs_cis: torch.Tensor,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_value: Optional[Tuple[torch.Tensor]] = None,
output_attentions: bool = False,
use_cache: bool = False,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
# Self Attention
residual = hidden_states
hidden_states = self.input_layernorm(hidden_states)
hidden_states, self_attn_weights, present_key_value = self.self_attn(
hidden_states=hidden_states,
freqs_cis=freqs_cis,
attention_mask=attention_mask,
position_ids=position_ids,
past_key_value=past_key_value,
output_attentions=output_attentions,
use_cache=use_cache,
)
hidden_states = residual + hidden_states
# Fully Connected
residual = hidden_states
hidden_states = self.post_attention_layernorm(hidden_states)
hidden_states = self.mlp(hidden_states)
hidden_states = residual + hidden_states
outputs = (hidden_states,)
if output_attentions:
outputs += (self_attn_weights,)
if use_cache:
outputs += (present_key_value,)
return outputs
class SapnousVisionEmbeddings(nn.Module):
"""Vision embeddings for multimodal support."""
def __init__(self, config):
super().__init__()
self.config = config
self.hidden_size = config.hidden_size
# Vision embedding layers
self.patch_embed = nn.Conv2d(3, self.hidden_size, kernel_size=16, stride=16)
self.cls_token = nn.Parameter(torch.zeros(1, 1, self.hidden_size))
self.pos_embed = nn.Parameter(torch.zeros(1, (224 // 16) ** 2 + 1, self.hidden_size))
# Layer normalization and dropout
self.norm = nn.LayerNorm(self.hidden_size, eps=config.rms_norm_eps)
self.dropout = nn.Dropout(config.attention_dropout)
def forward(self, pixel_values: torch.Tensor) -> torch.Tensor:
B = pixel_values.shape[0]
# Create patch embeddings
x = self.patch_embed(pixel_values)
x = x.flatten(2).transpose(1, 2) # B, N, C
# Add cls token and position embeddings
cls_tokens = self.cls_token.expand(B, -1, -1)
x = torch.cat((cls_tokens, x), dim=1)
x = x + self.pos_embed
# Apply normalization and dropout
x = self.norm(x)
x = self.dropout(x)
return x |