use tch::{IndexOp, Tensor, nn};
use crate::{
config::{Pool, ViTConfig},
layers::FeedForward,
models::vit::PatchEmbedding,
tensor::{merge_heads, num_patches, patch_dim, repeat_token, split_heads},
};
#[derive(Debug)]
struct ReAttention {
norm: nn::LayerNorm,
to_qkv: nn::Linear,
reattn_weights: Tensor,
reattn_norm: nn::LayerNorm,
to_out: nn::Linear,
heads: i64,
scale: f64,
dropout: f64,
}
impl ReAttention {
fn new(vs: &nn::Path, dim: i64, heads: i64, dim_head: i64, dropout: f64) -> Self {
let inner_dim = heads * dim_head;
let norm = nn::layer_norm(vs / "norm", vec![dim], Default::default());
let to_qkv = nn::linear(
vs / "to_qkv",
dim,
inner_dim * 3,
nn::LinearConfig {
bias: false,
..Default::default()
},
);
let reattn_weights = vs.var(
"reattn_weights",
&[heads, heads],
nn::Init::Randn {
mean: 0.0,
stdev: 1.0,
},
);
let reattn_norm = nn::layer_norm(vs / "reattn_norm", vec![heads], Default::default());
let to_out = nn::linear(vs / "to_out", inner_dim, dim, Default::default());
Self {
norm,
to_qkv,
reattn_weights,
reattn_norm,
to_out,
heads,
scale: (dim_head as f64).powf(-0.5),
dropout,
}
}
}
impl nn::ModuleT for ReAttention {
fn forward_t(&self, xs: &Tensor, train: bool) -> Tensor {
let qkv = xs.apply(&self.norm).apply(&self.to_qkv).chunk(3, -1);
let q = split_heads(&qkv[0], self.heads);
let k = split_heads(&qkv[1], self.heads);
let v = split_heads(&qkv[2], self.heads);
let attn = (q.matmul(&k.transpose(-1, -2)) * self.scale)
.softmax(-1, q.kind())
.dropout(self.dropout, train);
let attn = attn
.permute([0, 2, 3, 1])
.matmul(&self.reattn_weights)
.permute([0, 3, 1, 2]);
let attn = attn
.permute([0, 2, 3, 1])
.apply(&self.reattn_norm)
.permute([0, 3, 1, 2]);
merge_heads(&attn.matmul(&v))
.apply(&self.to_out)
.dropout(self.dropout, train)
}
}
#[derive(Debug)]
struct DeepViTLayer {
attention: ReAttention,
feed_forward: FeedForward,
}
impl DeepViTLayer {
fn new(vs: &nn::Path, dim: i64, heads: i64, dim_head: i64, mlp_dim: i64, dropout: f64) -> Self {
let attention = ReAttention::new(&(vs / "attention"), dim, heads, dim_head, dropout);
let feed_forward = FeedForward::new(&(vs / "feed_forward"), dim, mlp_dim, dropout);
Self {
attention,
feed_forward,
}
}
}
impl nn::ModuleT for DeepViTLayer {
fn forward_t(&self, xs: &Tensor, train: bool) -> Tensor {
let xs = xs + self.attention.forward_t(xs, train);
let ff_out = self.feed_forward.forward_t(&xs, train);
xs + ff_out
}
}
#[derive(Debug)]
struct DeepViTTransformer {
layers: Vec<DeepViTLayer>,
}
impl DeepViTTransformer {
fn new(
vs: &nn::Path,
dim: i64,
depth: usize,
heads: i64,
dim_head: i64,
mlp_dim: i64,
dropout: f64,
) -> Self {
let layers = (0..depth)
.map(|index| {
let name = format!("layer_{index}");
DeepViTLayer::new(
&(vs / name.as_str()),
dim,
heads,
dim_head,
mlp_dim,
dropout,
)
})
.collect();
Self { layers }
}
}
impl nn::ModuleT for DeepViTTransformer {
fn forward_t(&self, xs: &Tensor, train: bool) -> Tensor {
let mut xs = xs.shallow_clone();
for layer in &self.layers {
xs = layer.forward_t(&xs, train);
}
xs
}
}
#[derive(Debug)]
pub struct DeepViT {
patch_embedding: PatchEmbedding,
cls_token: Tensor,
pos_embedding: Tensor,
transformer: DeepViTTransformer,
mlp_norm: nn::LayerNorm,
mlp_head: nn::Linear,
pool: Pool,
emb_dropout: f64,
}
impl DeepViT {
pub fn new(vs: &nn::Path, config: ViTConfig) -> Self {
let image_height = config.image_size.height;
let image_width = config.image_size.width;
assert_eq!(
image_height, image_width,
"DeepViT expects square image size"
);
let patch_height = config.patch_size.height;
let patch_width = config.patch_size.width;
assert_eq!(
patch_height, patch_width,
"DeepViT expects square patch size"
);
let num_patches = num_patches(image_height, image_width, patch_height, patch_width);
let patch_dim = patch_dim(config.channels, patch_height, patch_width);
let patch_embedding = PatchEmbedding::new(
&(vs / "patch_embedding"),
patch_dim,
config.dim,
patch_height,
patch_width,
);
let pos_embedding = vs.var(
"pos_embedding",
&[1, num_patches + 1, config.dim],
nn::Init::Randn {
mean: 0.0,
stdev: 1.0,
},
);
let cls_token = vs.var(
"cls_token",
&[1, 1, config.dim],
nn::Init::Randn {
mean: 0.0,
stdev: 1.0,
},
);
let transformer = DeepViTTransformer::new(
&(vs / "transformer"),
config.dim,
config.depth,
config.heads,
config.dim_head,
config.mlp_dim,
config.dropout,
);
let mlp_norm = nn::layer_norm(vs / "mlp_norm", vec![config.dim], Default::default());
let mlp_head = nn::linear(
vs / "mlp_head",
config.dim,
config.num_classes,
Default::default(),
);
Self {
patch_embedding,
cls_token,
pos_embedding,
transformer,
mlp_norm,
mlp_head,
pool: config.pool,
emb_dropout: config.emb_dropout,
}
}
}
impl nn::ModuleT for DeepViT {
fn forward_t(&self, xs: &Tensor, train: bool) -> Tensor {
let mut xs = xs.apply(&self.patch_embedding);
let batch = xs.size()[0];
let tokens = xs.size()[1];
let cls_tokens = repeat_token(&self.cls_token, batch);
xs = Tensor::cat(&[cls_tokens, xs], 1);
xs = (xs + self.pos_embedding.i((.., 0..tokens + 1, ..))).dropout(self.emb_dropout, train);
let xs = self.transformer.forward_t(&xs, train);
let pooled = match self.pool {
Pool::Mean => xs.mean_dim(1, false, xs.kind()),
Pool::Cls => xs.i((.., 0)),
};
pooled.apply(&self.mlp_norm).apply(&self.mlp_head)
}
}