Add tracing to segment-anything (#777)

* Tracing support for segment-anything.

* More tracing.

* Handle the empty slice case.
This commit is contained in:
Laurent Mazare
2023-09-08 15:31:29 +01:00
committed by GitHub
parent e5703d2f56
commit 158ff3c609
6 changed files with 90 additions and 16 deletions

View File

@ -1,9 +1,10 @@
use candle::{DType, IndexOp, Result, Tensor};
use candle_nn::{layer_norm, LayerNorm, Linear, Module, VarBuilder};
use candle_nn::{layer_norm, LayerNorm, Module, VarBuilder};
#[derive(Debug)]
struct PatchEmbed {
proj: candle_nn::Conv2d,
span: tracing::Span,
}
impl PatchEmbed {
@ -21,23 +22,28 @@ impl PatchEmbed {
..Default::default()
};
let proj = candle_nn::conv2d(in_chans, embed_dim, k_size, cfg, vb.pp("proj"))?;
Ok(Self { proj })
let span = tracing::span!(tracing::Level::TRACE, "patch-embed");
Ok(Self { proj, span })
}
}
impl Module for PatchEmbed {
fn forward(&self, xs: &Tensor) -> Result<Tensor> {
let _enter = self.span.enter();
xs.apply(&self.proj)?.permute((0, 2, 3, 1))
}
}
#[derive(Debug)]
struct Attention {
qkv: Linear,
proj: Linear,
qkv: crate::Linear,
proj: crate::Linear,
num_heads: usize,
scale: f64,
rel_pos_hw: Option<(Tensor, Tensor)>,
span: tracing::Span,
span_rel_pos: tracing::Span,
span_softmax: tracing::Span,
}
impl Attention {
@ -49,6 +55,9 @@ impl Attention {
input_size: (usize, usize),
vb: VarBuilder,
) -> Result<Self> {
let span = tracing::span!(tracing::Level::TRACE, "attention");
let span_rel_pos = tracing::span!(tracing::Level::TRACE, "attn-rel-pos");
let span_softmax = tracing::span!(tracing::Level::TRACE, "attn-sm");
let qkv = crate::linear(vb.pp("qkv"), dim, dim * 3, qkv_bias)?;
let proj = crate::linear(vb.pp("proj"), dim, dim, true)?;
let head_dim = dim / num_heads;
@ -66,6 +75,9 @@ impl Attention {
num_heads,
scale,
rel_pos_hw,
span,
span_rel_pos,
span_softmax,
})
}
@ -126,6 +138,7 @@ fn get_rel_pos(q_size: usize, k_size: usize, rel_pos: &Tensor) -> Result<Tensor>
impl Module for Attention {
fn forward(&self, xs: &Tensor) -> Result<Tensor> {
let _enter = self.span.enter();
let (b, h, w, c) = xs.dims4()?;
let qkv = self
.qkv
@ -137,8 +150,14 @@ impl Module for Attention {
let k = qkv.i(1)?;
let v = qkv.i(2)?;
let attn = (&q * self.scale)?.matmul(&k.t()?)?;
let attn = self.add_decomposed_rel_pos(attn, &q, (h, w), (h, w))?;
let attn = candle_nn::ops::softmax_last_dim(&attn)?;
let attn = {
let _enter = self.span_rel_pos.enter();
self.add_decomposed_rel_pos(attn, &q, (h, w), (h, w))?
};
let attn = {
let _enter = self.span_softmax.enter();
candle_nn::ops::softmax_last_dim(&attn)?
};
let attn = attn.matmul(&v)?;
let attn = attn
.reshape((b, self.num_heads, h, w, c / self.num_heads))?
@ -155,6 +174,7 @@ struct Block {
norm2: LayerNorm,
mlp: crate::MlpBlock,
window_size: usize,
span: tracing::Span,
}
impl Block {
@ -183,12 +203,14 @@ impl Block {
vb.pp("attn"),
)?;
let mlp = crate::MlpBlock::new(dim, dim * 4, candle_nn::Activation::Gelu, vb.pp("mlp"))?;
let span = tracing::span!(tracing::Level::TRACE, "ie-block");
Ok(Self {
norm1,
attn,
norm2,
mlp,
window_size,
span,
})
}
}
@ -249,6 +271,7 @@ fn window_unpartition(
impl Module for Block {
fn forward(&self, xs: &Tensor) -> Result<Tensor> {
let _enter = self.span.enter();
let shortcut = xs;
let xs = self.norm1.forward(xs)?;
let hw = (xs.dim(1)?, xs.dim(2)?);
@ -277,6 +300,7 @@ pub struct ImageEncoderViT {
neck_conv2: candle_nn::Conv2d,
neck_ln2: crate::LayerNorm2d,
pos_embed: Option<Tensor>,
span: tracing::Span,
}
impl ImageEncoderViT {
@ -346,6 +370,7 @@ impl ImageEncoderViT {
} else {
None
};
let span = tracing::span!(tracing::Level::TRACE, "image-encoder-vit");
Ok(Self {
patch_embed,
blocks,
@ -354,12 +379,14 @@ impl ImageEncoderViT {
neck_conv2,
neck_ln2,
pos_embed,
span,
})
}
}
impl Module for ImageEncoderViT {
fn forward(&self, xs: &Tensor) -> Result<Tensor> {
let _enter = self.span.enter();
let xs = self.patch_embed.forward(xs)?;
let mut xs = match &self.pos_embed {
Some(pos_embed) => (xs + pos_embed)?,