Add SuperTransformer
This commit is contained in:
D-X-Y
@@ -6,236 +6,186 @@ from __future__ import print_function
|
||||
|
||||
import math
|
||||
from functools import partial
|
||||
from typing import Optional, Text
|
||||
from typing import Optional, Text, List
|
||||
|
||||
import torch
|
||||
import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
|
||||
import xlayers
|
||||
import spaces
|
||||
from xlayers import trunc_normal_
|
||||
from xlayers import super_core
|
||||
|
||||
|
||||
DEFAULT_NET_CONFIG = dict(
|
||||
__all__ = ["DefaultSearchSpace"]
|
||||
|
||||
|
||||
def _get_mul_specs(candidates, num):
|
||||
results = []
|
||||
for i in range(num):
|
||||
results.append(spaces.Categorical(*candidates))
|
||||
return results
|
||||
|
||||
|
||||
def _get_list_mul(num, multipler):
|
||||
results = []
|
||||
for i in range(1, num + 1):
|
||||
results.append(i * multipler)
|
||||
return results
|
||||
|
||||
|
||||
def _assert_types(x, expected_types):
|
||||
if not isinstance(x, expected_types):
|
||||
raise TypeError(
|
||||
"The type [{:}] is expected to be {:}.".format(type(x), expected_types)
|
||||
)
|
||||
|
||||
|
||||
_default_max_depth = 5
|
||||
DefaultSearchSpace = dict(
|
||||
d_feat=6,
|
||||
embed_dim=64,
|
||||
depth=5,
|
||||
num_heads=4,
|
||||
mlp_ratio=4.0,
|
||||
stem_dim=spaces.Categorical(*_get_list_mul(8, 16)),
|
||||
embed_dims=_get_mul_specs(_get_list_mul(8, 16), _default_max_depth),
|
||||
num_heads=_get_mul_specs((1, 2, 4, 8), _default_max_depth),
|
||||
mlp_hidden_multipliers=_get_mul_specs((0.5, 1, 2, 4, 8), _default_max_depth),
|
||||
qkv_bias=True,
|
||||
pos_drop=0.0,
|
||||
mlp_drop_rate=0.0,
|
||||
attn_drop_rate=0.0,
|
||||
drop_path_rate=0.0,
|
||||
other_drop=0.0,
|
||||
)
|
||||
|
||||
|
||||
# Real Model
|
||||
class SuperTransformer(super_core.SuperModule):
|
||||
"""The super model for transformer."""
|
||||
|
||||
|
||||
class Attention(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
dim,
|
||||
num_heads=8,
|
||||
qkv_bias=False,
|
||||
qk_scale=None,
|
||||
attn_drop=0.0,
|
||||
proj_drop=0.0,
|
||||
):
|
||||
super(Attention, self).__init__()
|
||||
self.num_heads = num_heads
|
||||
head_dim = dim // num_heads
|
||||
self.scale = qk_scale or math.sqrt(head_dim)
|
||||
|
||||
self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
|
||||
self.attn_drop = nn.Dropout(attn_drop)
|
||||
self.proj = nn.Linear(dim, dim)
|
||||
self.proj_drop = nn.Dropout(proj_drop)
|
||||
|
||||
def forward(self, x):
|
||||
B, N, C = x.shape
|
||||
qkv = (
|
||||
self.qkv(x)
|
||||
.reshape(B, N, 3, self.num_heads, C // self.num_heads)
|
||||
.permute(2, 0, 3, 1, 4)
|
||||
)
|
||||
q, k, v = (
|
||||
qkv[0],
|
||||
qkv[1],
|
||||
qkv[2],
|
||||
) # make torchscript happy (cannot use tensor as tuple)
|
||||
|
||||
attn = (q @ k.transpose(-2, -1)) * self.scale
|
||||
attn = attn.softmax(dim=-1)
|
||||
attn = self.attn_drop(attn)
|
||||
|
||||
x = (attn @ v).transpose(1, 2).reshape(B, N, C)
|
||||
x = self.proj(x)
|
||||
x = self.proj_drop(x)
|
||||
return x
|
||||
|
||||
|
||||
class Block(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
dim,
|
||||
num_heads,
|
||||
mlp_ratio=4.0,
|
||||
qkv_bias=False,
|
||||
qk_scale=None,
|
||||
attn_drop=0.0,
|
||||
mlp_drop=0.0,
|
||||
drop_path=0.0,
|
||||
act_layer=nn.GELU,
|
||||
norm_layer=nn.LayerNorm,
|
||||
):
|
||||
super(Block, self).__init__()
|
||||
self.norm1 = norm_layer(dim)
|
||||
self.attn = Attention(
|
||||
dim,
|
||||
num_heads=num_heads,
|
||||
qkv_bias=qkv_bias,
|
||||
qk_scale=qk_scale,
|
||||
attn_drop=attn_drop,
|
||||
proj_drop=mlp_drop,
|
||||
)
|
||||
# NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
|
||||
self.drop_path = (
|
||||
xlayers.DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
|
||||
)
|
||||
self.norm2 = norm_layer(dim)
|
||||
mlp_hidden_dim = int(dim * mlp_ratio)
|
||||
self.mlp = xlayers.MLP(
|
||||
in_features=dim,
|
||||
hidden_features=mlp_hidden_dim,
|
||||
act_layer=act_layer,
|
||||
drop=mlp_drop,
|
||||
)
|
||||
|
||||
def forward(self, x):
|
||||
x = x + self.drop_path(self.attn(self.norm1(x)))
|
||||
x = x + self.drop_path(self.mlp(self.norm2(x)))
|
||||
return x
|
||||
|
||||
|
||||
class SimpleEmbed(nn.Module):
|
||||
def __init__(self, d_feat, embed_dim):
|
||||
super(SimpleEmbed, self).__init__()
|
||||
self.d_feat = d_feat
|
||||
self.embed_dim = embed_dim
|
||||
self.proj = nn.Linear(d_feat, embed_dim)
|
||||
|
||||
def forward(self, x):
|
||||
x = x.reshape(len(x), self.d_feat, -1) # [N, F*T] -> [N, F, T]
|
||||
x = x.permute(0, 2, 1) # [N, F, T] -> [N, T, F]
|
||||
out = self.proj(x) * math.sqrt(self.embed_dim)
|
||||
return out
|
||||
|
||||
|
||||
class TransformerModel(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
d_feat: int = 6,
|
||||
embed_dim: int = 64,
|
||||
depth: int = 4,
|
||||
num_heads: int = 4,
|
||||
mlp_ratio: float = 4.0,
|
||||
qkv_bias: bool = True,
|
||||
qk_scale: Optional[float] = None,
|
||||
pos_drop: float = 0.0,
|
||||
mlp_drop_rate: float = 0.0,
|
||||
attn_drop_rate: float = 0.0,
|
||||
drop_path_rate: float = 0.0,
|
||||
norm_layer: Optional[nn.Module] = None,
|
||||
stem_dim: super_core.IntSpaceType = DefaultSearchSpace["stem_dim"],
|
||||
embed_dims: List[super_core.IntSpaceType] = DefaultSearchSpace["embed_dims"],
|
||||
num_heads: List[super_core.IntSpaceType] = DefaultSearchSpace["num_heads"],
|
||||
mlp_hidden_multipliers: List[super_core.IntSpaceType] = DefaultSearchSpace[
|
||||
"mlp_hidden_multipliers"
|
||||
],
|
||||
qkv_bias: bool = DefaultSearchSpace["qkv_bias"],
|
||||
pos_drop: float = DefaultSearchSpace["pos_drop"],
|
||||
other_drop: float = DefaultSearchSpace["other_drop"],
|
||||
max_seq_len: int = 65,
|
||||
):
|
||||
"""
|
||||
Args:
|
||||
d_feat (int, tuple): input image size
|
||||
embed_dim (int): embedding dimension
|
||||
depth (int): depth of transformer
|
||||
num_heads (int): number of attention heads
|
||||
mlp_ratio (int): ratio of mlp hidden dim to embedding dim
|
||||
qkv_bias (bool): enable bias for qkv if True
|
||||
qk_scale (float): override default qk scale of head_dim ** -0.5 if set
|
||||
pos_drop (float): dropout rate for the positional embedding
|
||||
mlp_drop_rate (float): the dropout rate for MLP layers in a block
|
||||
attn_drop_rate (float): attention dropout rate
|
||||
drop_path_rate (float): stochastic depth rate
|
||||
norm_layer: (nn.Module): normalization layer
|
||||
"""
|
||||
super(TransformerModel, self).__init__()
|
||||
self.embed_dim = embed_dim
|
||||
self.num_features = embed_dim
|
||||
norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6)
|
||||
super(SuperTransformer, self).__init__()
|
||||
self._embed_dims = embed_dims
|
||||
self._stem_dim = stem_dim
|
||||
self._num_heads = num_heads
|
||||
self._mlp_hidden_multipliers = mlp_hidden_multipliers
|
||||
|
||||
self.input_embed = SimpleEmbed(d_feat, embed_dim=embed_dim)
|
||||
|
||||
self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
|
||||
self.pos_embed = xlayers.PositionalEncoder(
|
||||
d_model=embed_dim, max_seq_len=max_seq_len, dropout=pos_drop
|
||||
# the stem part
|
||||
self.input_embed = super_core.SuperAlphaEBDv1(d_feat, stem_dim)
|
||||
self.cls_token = nn.Parameter(torch.zeros(1, 1, self.stem_dim))
|
||||
self.pos_embed = super_core.SuperPositionalEncoder(
|
||||
d_model=stem_dim, max_seq_len=max_seq_len, dropout=pos_drop
|
||||
)
|
||||
|
||||
dpr = [
|
||||
x.item() for x in torch.linspace(0, drop_path_rate, depth)
|
||||
] # stochastic depth decay rule
|
||||
self.blocks = nn.ModuleList(
|
||||
[
|
||||
Block(
|
||||
dim=embed_dim,
|
||||
num_heads=num_heads,
|
||||
mlp_ratio=mlp_ratio,
|
||||
qkv_bias=qkv_bias,
|
||||
qk_scale=qk_scale,
|
||||
attn_drop=attn_drop_rate,
|
||||
mlp_drop=mlp_drop_rate,
|
||||
drop_path=dpr[i],
|
||||
norm_layer=norm_layer,
|
||||
)
|
||||
for i in range(depth)
|
||||
]
|
||||
# build the transformer encode layers -->> check params
|
||||
_assert_types(embed_dims, (tuple, list))
|
||||
_assert_types(num_heads, (tuple, list))
|
||||
_assert_types(mlp_hidden_multipliers, (tuple, list))
|
||||
num_layers = len(embed_dims)
|
||||
assert (
|
||||
num_layers == len(num_heads) == len(mlp_hidden_multipliers)
|
||||
), "{:} vs {:} vs {:}".format(
|
||||
num_layers, len(num_heads), len(mlp_hidden_multipliers)
|
||||
)
|
||||
self.norm = norm_layer(embed_dim)
|
||||
# build the transformer encode layers -->> backbone
|
||||
layers, input_dim = [], stem_dim
|
||||
for embed_dim, num_head, mlp_hidden_multiplier in zip(
|
||||
embed_dims, num_heads, mlp_hidden_multipliers
|
||||
):
|
||||
layer = super_core.SuperTransformerEncoderLayer(
|
||||
input_dim,
|
||||
embed_dim,
|
||||
num_head,
|
||||
qkv_bias,
|
||||
mlp_hidden_multiplier,
|
||||
other_drop,
|
||||
)
|
||||
layers.append(layer)
|
||||
input_dim = embed_dim
|
||||
self.backbone = super_core.SuperSequential(*layers)
|
||||
|
||||
# regression head
|
||||
self.head = nn.Linear(self.num_features, 1)
|
||||
|
||||
xlayers.trunc_normal_(self.cls_token, std=0.02)
|
||||
# the regression head
|
||||
self.head = super_core.SuperLinear(self._embed_dims[-1], 1)
|
||||
trunc_normal_(self.cls_token, std=0.02)
|
||||
self.apply(self._init_weights)
|
||||
|
||||
@property
|
||||
def stem_dim(self):
|
||||
return spaces.get_max(self._stem_dim)
|
||||
|
||||
@property
|
||||
def abstract_search_space(self):
|
||||
root_node = spaces.VirtualNode(id(self))
|
||||
xdict = dict(
|
||||
input_embed=self.input_embed.abstract_search_space,
|
||||
pos_embed=self.pos_embed.abstract_search_space,
|
||||
backbone=self.backbone.abstract_search_space,
|
||||
head=self.head.abstract_search_space,
|
||||
)
|
||||
if not spaces.is_determined(self._stem_dim):
|
||||
root_node.append("_stem_dim", self._stem_dim.abstract(reuse_last=True))
|
||||
for key, space in xdict.items():
|
||||
if not spaces.is_determined(space):
|
||||
root_node.append(key, space)
|
||||
return root_node
|
||||
|
||||
def apply_candidate(self, abstract_child: spaces.VirtualNode):
|
||||
super(SuperTransformer, self).apply_candidate(abstract_child)
|
||||
xkeys = ("input_embed", "pos_embed", "backbone", "head")
|
||||
for key in xkeys:
|
||||
if key in abstract_child:
|
||||
getattr(self, key).apply_candidate(abstract_child[key])
|
||||
|
||||
def _init_weights(self, m):
|
||||
if isinstance(m, nn.Linear):
|
||||
xlayers.trunc_normal_(m.weight, std=0.02)
|
||||
trunc_normal_(m.weight, std=0.02)
|
||||
if isinstance(m, nn.Linear) and m.bias is not None:
|
||||
nn.init.constant_(m.bias, 0)
|
||||
elif isinstance(m, nn.LayerNorm):
|
||||
nn.init.constant_(m.bias, 0)
|
||||
elif isinstance(m, super_core.SuperLinear):
|
||||
trunc_normal_(m._super_weight, std=0.02)
|
||||
if m._super_bias is not None:
|
||||
nn.init.constant_(m._super_bias, 0)
|
||||
elif isinstance(m, super_core.SuperLayerNorm1D):
|
||||
nn.init.constant_(m.weight, 1.0)
|
||||
nn.init.constant_(m.bias, 0)
|
||||
|
||||
def forward_features(self, x):
|
||||
batch, flatten_size = x.shape
|
||||
feats = self.input_embed(x) # batch * 60 * 64
|
||||
|
||||
cls_tokens = self.cls_token.expand(
|
||||
batch, -1, -1
|
||||
) # stole cls_tokens impl from Phil Wang, thanks
|
||||
def forward_candidate(self, input: torch.Tensor) -> torch.Tensor:
|
||||
batch, flatten_size = input.shape
|
||||
feats = self.input_embed(input) # batch * 60 * 64
|
||||
if not spaces.is_determined(self._stem_dim):
|
||||
stem_dim = self.abstract_child["_stem_dim"].value
|
||||
else:
|
||||
stem_dim = spaces.get_determined_value(self._stem_dim)
|
||||
cls_tokens = self.cls_token.expand(batch, -1, -1)
|
||||
cls_tokens = F.interpolate(cls_tokens, size=(stem_dim), mode="linear", align_corners=True)
|
||||
feats_w_ct = torch.cat((cls_tokens, feats), dim=1)
|
||||
feats_w_tp = self.pos_embed(feats_w_ct)
|
||||
xfeats = self.backbone(feats_w_tp)
|
||||
xfeats = xfeats[:, 0, :] # use the feature for the first token
|
||||
predicts = self.head(xfeats).squeeze(-1)
|
||||
return predicts
|
||||
|
||||
xfeats = feats_w_tp
|
||||
for block in self.blocks:
|
||||
xfeats = block(xfeats)
|
||||
|
||||
xfeats = self.norm(xfeats)[:, 0]
|
||||
return xfeats
|
||||
|
||||
def forward(self, x):
|
||||
feats = self.forward_features(x)
|
||||
predicts = self.head(feats).squeeze(-1)
|
||||
def forward_raw(self, input: torch.Tensor) -> torch.Tensor:
|
||||
batch, flatten_size = input.shape
|
||||
feats = self.input_embed(input) # batch * 60 * 64
|
||||
cls_tokens = self.cls_token.expand(batch, -1, -1)
|
||||
feats_w_ct = torch.cat((cls_tokens, feats), dim=1)
|
||||
feats_w_tp = self.pos_embed(feats_w_ct)
|
||||
xfeats = self.backbone(feats_w_tp)
|
||||
xfeats = xfeats[:, 0, :] # use the feature for the first token
|
||||
predicts = self.head(xfeats).squeeze(-1)
|
||||
return predicts
|
||||
|
||||
|
||||
def get_transformer(config):
|
||||
if config is None:
|
||||
return SuperTransformer(6)
|
||||
if not isinstance(config, dict):
|
||||
raise ValueError("Invalid Configuration: {:}".format(config))
|
||||
name = config.get("name", "basic")
|
||||
|
||||
Reference in New Issue
Block a user