tckm.yan/aha
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update fmt

Browse Source
This commit is contained in:
jhqxxx
2026-03-14 20:06:20 +08:00
parent 0c6d4d01d9
commit 21caee0522
13 changed files with 761 additions and 697 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Cargo.lock
Generated
+698 -639
View File
File diff suppressed because it is too large Load Diff
rustfmt.toml
-9
View File
@@ -1,9 +0,0 @@
edition = "2024"
style_edition = "2024"
reorder_imports = true
group_imports = "StdExternalCrate"
where_single_line = true
trailing_comma = "Vertical"
overflow_delimited_expr = true
format_code_in_doc_comments = true
normalize_comments = true
src/models/deepseek_ocr/model.rs
+4 -2
View File
@@ -492,9 +492,11 @@ impl ImageEncoderViT {
} }
pub fn forward(&self, xs: &Tensor) -> Result<Tensor> { pub fn forward(&self, xs: &Tensor) -> Result<Tensor> {
let mut x = self.patch_embed.forward(xs)?; let mut x = self.patch_embed.forward(xs)?;
if self.pos_embed.is_some() { // if self.pos_embed.is_some() {
if let Some(pos_emb) = &self.pos_embed {
let dim1 = x.dim(1)?; let dim1 = x.dim(1)?;
let pos = self.get_abs_pos_sam(self.pos_embed.as_ref().unwrap(), dim1)?; // let pos = self.get_abs_pos_sam(self.pos_embed.as_ref().unwrap(), dim1)?;
let pos = self.get_abs_pos_sam(pos_emb, dim1)?;
x = x.broadcast_add(&pos)?; x = x.broadcast_add(&pos)?;
} }
for blk in &self.blocks { for blk in &self.blocks {
src/models/fun_asr_nano/model.rs
+8 -15
View File
@@ -268,19 +268,15 @@ impl EncoderLayerSANM {
self.self_attn self.self_attn
.forward(&xs, mask, mask_shfit_chunk, mask_att_chunk_encoder)?; .forward(&xs, mask, mask_shfit_chunk, mask_att_chunk_encoder)?;
let x_concat = Tensor::cat(&[&xs, &attn], D::Minus1)?; let x_concat = Tensor::cat(&[&xs, &attn], D::Minus1)?;
if self.in_dim == self.hidden_dim { if self.in_dim == self.hidden_dim
let x_concat = self && let Some(concat_linear) = &self.concat_linear
.concat_linear {
.as_ref() let x_concat = concat_linear
.unwrap()
.forward(&x_concat)? .forward(&x_concat)?
.affine(stoch_layer_coeff, 0.0)?; .affine(stoch_layer_coeff, 0.0)?;
xs = residual.add(&x_concat)?; xs = residual.add(&x_concat)?;
} else { } else if let Some(concat_linear) = &self.concat_linear {
xs = self xs = concat_linear
.concat_linear
.as_ref()
.unwrap()
.forward(&x_concat)? .forward(&x_concat)?
.affine(stoch_layer_coeff, 0.0)?; .affine(stoch_layer_coeff, 0.0)?;
} }
@@ -496,13 +492,10 @@ impl AdaptorEncoderLayer {
} else { } else {
xs.clone() xs.clone()
}; };
if self.concat_linear.is_some() { if let Some(concat_linear) = &self.concat_linear {
let attn = self.self_attn.forward(&xs, None, None, mask, false)?; let attn = self.self_attn.forward(&xs, None, None, mask, false)?;
let x_concat = Tensor::cat(&[&xs, &attn], D::Minus1)?; let x_concat = Tensor::cat(&[&xs, &attn], D::Minus1)?;
let x_concat = self let x_concat = concat_linear
.concat_linear
.as_ref()
.unwrap()
.forward(&x_concat)? .forward(&x_concat)?
.affine(stoch_layer_coeff, 0.0)?; .affine(stoch_layer_coeff, 0.0)?;
xs = residual.add(&x_concat)?; xs = residual.add(&x_concat)?;
src/models/paddleocr_vl/model.rs
+5 -3
View File
@@ -707,11 +707,13 @@ impl PaddleOCRVLModel {
self.rope_deltas = Some(rope_deltas); self.rope_deltas = Some(rope_deltas);
} else { } else {
let (bs, seq_len, _) = inputs_embeds.dims3()?; let (bs, seq_len, _) = inputs_embeds.dims3()?;
let delta = if let Some(cache_position) = cache_position { let delta = if let Some(cache_position) = cache_position
&& let Some(rope_deltas) = &self.rope_deltas
{
cache_position cache_position
.i(0)? .i(0)?
.to_dtype(self.rope_deltas.as_ref().unwrap().dtype())? .to_dtype(rope_deltas.dtype())?
.broadcast_add(self.rope_deltas.as_ref().unwrap())? .broadcast_add(rope_deltas)?
.contiguous()? .contiguous()?
.to_dtype(candle_core::DType::U32)? .to_dtype(candle_core::DType::U32)?
} else { } else {
src/models/qwen2_5vl/model.rs
+5 -3
View File
@@ -1057,11 +1057,13 @@ impl Qwen2_5VLModel {
self.rope_deltas = Some(rope_deltas); self.rope_deltas = Some(rope_deltas);
} else { } else {
let (bs, seq_len, _) = inputs_embeds.dims3()?; let (bs, seq_len, _) = inputs_embeds.dims3()?;
let delta = if let Some(cache_position) = cache_position { let delta = if let Some(cache_position) = cache_position
&& let Some(rope_deltas) = &self.rope_deltas
{
cache_position cache_position
.i(0)? .i(0)?
.to_dtype(self.rope_deltas.as_ref().unwrap().dtype())? .to_dtype(rope_deltas.dtype())?
.broadcast_add(self.rope_deltas.as_ref().unwrap())? .broadcast_add(rope_deltas)?
.contiguous()? .contiguous()?
.to_dtype(candle_core::DType::U32)? .to_dtype(candle_core::DType::U32)?
} else { } else {
src/models/qwen3_5/model.rs
+8 -8
View File
@@ -1316,26 +1316,26 @@ impl Qwen3_5Model {
video_grid_thw: Option<&Tensor>, video_grid_thw: Option<&Tensor>,
seqlen_offset: usize, seqlen_offset: usize,
) -> Result<Tensor> { ) -> Result<Tensor> {
let position_ids = if self.rope_deltas.is_none() { let position_ids = if let Some(rope_deltas) = &self.rope_deltas {
let (position_ids, rope_deltas) =
self.get_rope_index(input_ids, image_grid_thw, video_grid_thw, None)?;
self.rope_deltas = Some(rope_deltas);
position_ids
} else {
let (bs, seq_len, _) = inputs_embeds.dims3()?; let (bs, seq_len, _) = inputs_embeds.dims3()?;
Tensor::arange( Tensor::arange(
seqlen_offset as i64, seqlen_offset as i64,
(seqlen_offset + seq_len) as i64, (seqlen_offset + seq_len) as i64,
input_ids.device(), input_ids.device(),
)? )?
.to_dtype(self.rope_deltas.as_ref().unwrap().dtype())? .to_dtype(rope_deltas.dtype())?
.unsqueeze(0)? .unsqueeze(0)?
.broadcast_as((bs, seq_len))? .broadcast_as((bs, seq_len))?
.broadcast_add(self.rope_deltas.as_ref().unwrap())? .broadcast_add(rope_deltas)?
.unsqueeze(0)? .unsqueeze(0)?
.broadcast_as((3, bs, seq_len))? .broadcast_as((3, bs, seq_len))?
.contiguous()? .contiguous()?
.to_dtype(candle_core::DType::U32)? .to_dtype(candle_core::DType::U32)?
} else {
let (position_ids, rope_deltas) =
self.get_rope_index(input_ids, image_grid_thw, video_grid_thw, None)?;
self.rope_deltas = Some(rope_deltas);
position_ids
}; };
Ok(position_ids) Ok(position_ids)
} }
src/models/qwen3vl/model.rs
+5 -3
View File
@@ -1013,11 +1013,13 @@ impl Qwen3VLModel {
self.rope_deltas = Some(rope_deltas); self.rope_deltas = Some(rope_deltas);
} else { } else {
let (bs, seq_len, _) = inputs_embeds.dims3()?; let (bs, seq_len, _) = inputs_embeds.dims3()?;
let delta = if let Some(cache_position) = cache_position { let delta = if let Some(cache_position) = cache_position
&& let Some(rope_deltas) = &self.rope_deltas
{
cache_position cache_position
.i(0)? .i(0)?
.to_dtype(self.rope_deltas.as_ref().unwrap().dtype())? .to_dtype(rope_deltas.dtype())?
.broadcast_add(self.rope_deltas.as_ref().unwrap())? .broadcast_add(rope_deltas)?
.contiguous()? .contiguous()?
.to_dtype(candle_core::DType::U32)? .to_dtype(candle_core::DType::U32)?
} else { } else {
src/models/qwen3vl/processor.rs
+3
View File
@@ -9,6 +9,9 @@ use candle_core::{DType, Device, IndexOp, Shape, Tensor};
use ffmpeg_next as ffmpeg; use ffmpeg_next as ffmpeg;
use image::DynamicImage; use image::DynamicImage;
#[cfg(feature = "ffmpeg")]
use anyhow::anyhow;
#[cfg(feature = "ffmpeg")] #[cfg(feature = "ffmpeg")]
use crate::utils::video_utils::video_smart_resize; use crate::utils::video_utils::video_smart_resize;
use crate::{ use crate::{
src/models/rmbg2_0/model.rs
+8 -5
View File
@@ -63,10 +63,10 @@ impl PatchEmbed {
xs = xs.pad_with_zeros(2, 0, self.patch_size - h % self.patch_size)?; xs = xs.pad_with_zeros(2, 0, self.patch_size - h % self.patch_size)?;
} }
xs = self.proj.forward(&xs)?; xs = self.proj.forward(&xs)?;
if self.norm.is_some() { if let Some(norm) = &self.norm {
let (_, _, ph, pw) = xs.dims4()?; let (_, _, ph, pw) = xs.dims4()?;
xs = xs.flatten_from(2)?.transpose(1, 2)?; xs = xs.flatten_from(2)?.transpose(1, 2)?;
xs = self.norm.as_ref().unwrap().forward(&xs)?; xs = norm.forward(&xs)?;
xs = xs.transpose(1, 2)?.reshape(((), self.embed_dim, ph, pw))?; xs = xs.transpose(1, 2)?.reshape(((), self.embed_dim, ph, pw))?;
} }
Ok(xs) Ok(xs)
@@ -1018,9 +1018,12 @@ impl BasicDecBlk {
pub fn new(vb: VarBuilder, in_c: usize, out_c: usize) -> Result<Self> { pub fn new(vb: VarBuilder, in_c: usize, out_c: usize) -> Result<Self> {
let inter_channels = 64; let inter_channels = 64;
let conv_in = get_conv2d(vb.pp("conv_in"), in_c, inter_channels, 3, 1, 1, 1, 1, true)?; let conv_in = get_conv2d(vb.pp("conv_in"), in_c, inter_channels, 3, 1, 1, 1, 1, true)?;
let dec_att = ASPPDeformable::new(vb.pp("dec_att"), inter_channels, inter_channels, vec![ let dec_att = ASPPDeformable::new(
1, 3, 7, vb.pp("dec_att"),
])?; inter_channels,
inter_channels,
vec![1, 3, 7],
)?;
let conv_out = get_conv2d( let conv_out = get_conv2d(
vb.pp("conv_out"), vb.pp("conv_out"),
inter_channels, inter_channels,
src/position_embed/sinusoidal_pe.rs
+3 -3
View File
@@ -33,11 +33,11 @@ impl SinusoidalPositionEncoderCat {
device, device,
)? )?
.reshape((seq_len, 1))?; // (seq_len, 1) .reshape((seq_len, 1))?; // (seq_len, 1)
let inv_freq = if self.inv_freq.is_none() { let inv_freq = if let Some(inv_freq) = &self.inv_freq {
inv_freq.clone()
} else {
let inv_freq = compute_default_rope_parameters(head_dim, 10000.0); let inv_freq = compute_default_rope_parameters(head_dim, 10000.0);
Tensor::from_slice(&inv_freq, (1, inv_freq.len()), device)? Tensor::from_slice(&inv_freq, (1, inv_freq.len()), device)?
} else {
self.inv_freq.as_ref().unwrap().clone()
}; };
let freqs = positions.matmul(&inv_freq)?; // (seq_len, dim / 2) let freqs = positions.matmul(&inv_freq)?; // (seq_len, dim / 2)
let sin = freqs.sin()?; let sin = freqs.sin()?;
tests/test_qwen3_5.rs
+8 -1
View File
@@ -20,9 +20,16 @@ fn qwen3_5_generate() -> Result<()> {
{ {
"role": "user", "role": "user",
"content": [ "content": [
{
"type": "image",
"image_url":
{
"url": "file:///home/jhq/Downloads/gougou1.jpg"
}
},
{ {
"type": "text", "type": "text",
"text": "你好啊，你是谁" "text": "描述这张图片."
} }
] ]
} }
tests/test_qwen3vl.rs
+5 -5
View File
@@ -59,7 +59,7 @@ fn qwen3vl_thinking_generate() -> Result<()> {
#[test] #[test]
fn qwen3vl_generate() -> Result<()> { fn qwen3vl_generate() -> Result<()> {
// test with cuda: RUST_BACKTRACE=1 cargo test -F cuda --test test_qwen3vl qwen3vl_generate -r -- --nocapture // test with cuda: RUST_BACKTRACE=1 cargo test -F cuda,ffmpeg --test test_qwen3vl qwen3vl_generate -r -- --nocapture
let save_dir = let save_dir =
aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?; aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
@@ -73,15 +73,15 @@ fn qwen3vl_generate() -> Result<()> {
"role": "user", "role": "user",
"content": [ "content": [
{ {
"type": "image", "type": "video",
"image_url": "video_url":
{ {
"url": "file:///home/jhq/Downloads/gougou1.jpg" "url": "./assets/video/video_test.mp4"
} }
}, },
{ {
"type": "text", "type": "text",
"text": "描述这张图片." "text": "视频中发生了什么？"
} }
] ]
} }