Ross Wightman
2 years ago
parent
4c8bb295ab
commit
7eb7e73216
@ -1,143 +0,0 @@
|
||||
from typing import List
|
||||
|
||||
import torch
|
||||
import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
|
||||
from .helpers import to_2tuple
|
||||
from .weight_init import trunc_normal_
|
||||
|
||||
|
||||
def rel_logits_1d(q, rel_k, permute_mask: List[int]):
|
||||
""" Compute relative logits along one dimension
|
||||
|
||||
As per: https://gist.github.com/aravindsrinivas/56359b79f0ce4449bcb04ab4b56a57a2
|
||||
Originally from: `Attention Augmented Convolutional Networks` - https://arxiv.org/abs/1904.09925
|
||||
|
||||
Args:
|
||||
q: (batch, heads, height, width, dim)
|
||||
rel_k: (2 * width - 1, dim)
|
||||
permute_mask: permute output dim according to this
|
||||
"""
|
||||
B, H, W, dim = q.shape
|
||||
x = (q @ rel_k.transpose(-1, -2))
|
||||
x = x.reshape(-1, W, 2 * W -1)
|
||||
|
||||
# pad to shift from relative to absolute indexing
|
||||
x_pad = F.pad(x, [0, 1]).flatten(1)
|
||||
x_pad = F.pad(x_pad, [0, W - 1])
|
||||
|
||||
# reshape and slice out the padded elements
|
||||
x_pad = x_pad.reshape(-1, W + 1, 2 * W - 1)
|
||||
x = x_pad[:, :W, W - 1:]
|
||||
|
||||
# reshape and tile
|
||||
x = x.reshape(B, H, 1, W, W).expand(-1, -1, H, -1, -1)
|
||||
return x.permute(permute_mask)
|
||||
|
||||
|
||||
class PosEmbedRel(nn.Module):
|
||||
""" Relative Position Embedding
|
||||
As per: https://gist.github.com/aravindsrinivas/56359b79f0ce4449bcb04ab4b56a57a2
|
||||
Originally from: `Attention Augmented Convolutional Networks` - https://arxiv.org/abs/1904.09925
|
||||
"""
|
||||
def __init__(self, feat_size, dim_head, scale):
|
||||
super().__init__()
|
||||
self.height, self.width = to_2tuple(feat_size)
|
||||
self.dim_head = dim_head
|
||||
self.scale = scale
|
||||
self.height_rel = nn.Parameter(torch.randn(self.height * 2 - 1, dim_head) * self.scale)
|
||||
self.width_rel = nn.Parameter(torch.randn(self.width * 2 - 1, dim_head) * self.scale)
|
||||
|
||||
def forward(self, q):
|
||||
B, num_heads, HW, _ = q.shape
|
||||
|
||||
# relative logits in width dimension.
|
||||
q = q.reshape(B * num_heads, self.height, self.width, -1)
|
||||
rel_logits_w = rel_logits_1d(q, self.width_rel, permute_mask=(0, 1, 3, 2, 4))
|
||||
|
||||
# relative logits in height dimension.
|
||||
q = q.transpose(1, 2)
|
||||
rel_logits_h = rel_logits_1d(q, self.height_rel, permute_mask=(0, 3, 1, 4, 2))
|
||||
|
||||
rel_logits = rel_logits_h + rel_logits_w
|
||||
rel_logits = rel_logits.reshape(B, num_heads, HW, HW)
|
||||
return rel_logits
|
||||
|
||||
|
||||
class BottleneckAttn(nn.Module):
|
||||
""" Bottleneck Attention
|
||||
Paper: `Bottleneck Transformers for Visual Recognition` - https://arxiv.org/abs/2101.11605
|
||||
"""
|
||||
def __init__(self, dim, dim_out=None, feat_size=None, stride=1, num_heads=4, qkv_bias=False):
|
||||
super().__init__()
|
||||
assert feat_size is not None, 'A concrete feature size matching expected input (H, W) is required'
|
||||
dim_out = dim_out or dim
|
||||
assert dim_out % num_heads == 0
|
||||
self.num_heads = num_heads
|
||||
self.dim_out = dim_out
|
||||
self.dim_head = dim_out // num_heads
|
||||
self.scale = self.dim_head ** -0.5
|
||||
|
||||
self.qkv = nn.Conv2d(dim, self.dim_out * 3, 1, bias=qkv_bias)
|
||||
|
||||
# NOTE I'm only supporting relative pos embedding for now
|
||||
self.pos_embed = PosEmbedRel(feat_size, dim_head=self.dim_head, scale=self.scale)
|
||||
|
||||
self.pool = nn.AvgPool2d(2, 2) if stride == 2 else nn.Identity()
|
||||
|
||||
self.reset_parameters()
|
||||
|
||||
def reset_parameters(self):
|
||||
trunc_normal_(self.qkv.weight, std=self.qkv.weight.shape[1] ** -0.5)
|
||||
trunc_normal_(self.pos_embed.height_rel, std=self.scale)
|
||||
trunc_normal_(self.pos_embed.width_rel, std=self.scale)
|
||||
|
||||
def forward(self, x):
|
||||
B, C, H, W = x.shape
|
||||
assert H == self.pos_embed.height
|
||||
assert W == self.pos_embed.width
|
||||
|
||||
x = self.qkv(x) # B, 3 * num_heads * dim_head, H, W
|
||||
x = x.reshape(B, -1, self.dim_head, H * W).transpose(-1, -2)
|
||||
q, k, v = torch.split(x, self.num_heads, dim=1)
|
||||
|
||||
attn_logits = (q @ k.transpose(-1, -2)) * self.scale
|
||||
attn_logits = attn_logits + self.pos_embed(q) # B, num_heads, H * W, H * W
|
||||
|
||||
attn_out = attn_logits.softmax(dim=-1)
|
||||
attn_out = (attn_out @ v).transpose(-1, -2).reshape(B, self.dim_out, H, W) # B, dim_out, H, W
|
||||
attn_out = self.pool(attn_out)
|
||||
return attn_out
|
||||
|
||||
|
||||
class PoolingAttention(nn.Module):
|
||||
def __init__(self, in_features: int, attention_features: int, segments: int, max_pool_kernel: int):
|
||||
super(PoolingAttention, self).__init__()
|
||||
self.attn = nn.Linear(in_features, attention_features * 5)
|
||||
self.segments = segments
|
||||
self.max_pool_kernel = max_pool_kernel
|
||||
|
||||
def forward(self, inp: torch.Tensor): # Shape: [Batch, Sequence, Features]
|
||||
batch, sequence, features = inp.size()
|
||||
assert sequence % self.segments == 0
|
||||
|
||||
qry, key, val, seg, loc = self.attn(inp).chunk(5, 2) # 5x Shape: [Batch, Sequence, AttentionFeatures]
|
||||
|
||||
aggregated = qry.mean(1, keepdim=True) # Shape: [Batch, AttentionFeatures]
|
||||
aggregated = torch.einsum("ba,bsa->bs", aggregated, key) # Shape: [Batch, Sequence]
|
||||
aggregated = F.softmax(aggregated, 1)
|
||||
aggregated = torch.einsum("bs,bsa,bza->bza", aggregated, val,
|
||||
qry) # Shape: [Batch, Sequence, AttentionFeatures]
|
||||
|
||||
pooled_sequence = sequence // self.segments
|
||||
segment_max_pooled = seg.view(batch, pooled_sequence, self.segments, -1)
|
||||
segment_max_pooled = segment_max_pooled.max(2, keepdim=True) # Shape: [Batch, PooledSequence, 1, AttentionFeatures]
|
||||
segment_max_pooled = segment_max_pooled * qry.view(batch, pooled_sequence, self.segments, -1) # Shape: [Batch, PooledSequence, PoolSize, AttentionFeatures]
|
||||
segment_max_pooled = segment_max_pooled.view(batch, sequence, -1) # Shape: [Batch, Sequence, AttentionFeatures]
|
||||
|
||||
loc = loc.transpose(1, 2) # Shape: [Batch, AttentionFeatures, Sequence]
|
||||
local_max_pooled = F.max_pool1d(loc, self.max_pool_kernel, 1, self.max_pool_kernel // 2)
|
||||
local_max_pooled = local_max_pooled.transpose(1, 2) # Shape: [Batch, Sequence, AttentionFeatures]
|
||||
|
||||
return aggregated + segment_max_pooled + local_max_pooled
|
||||
Loading…
Reference in new issue