onnx: fix pad, unsqueeze (#2317)

* onnx: fix pad, unsqueeze

both implementations have off-by-one errors:
- Pad 'reflect' cycle for eg `dim==3` is `[0,1,2,1]` which has length of
  4 (or `dim*2 - 2`) not 5 (current code `dim*2 - 1`)
- Unsqueeze(-1) for tensor with `dim==3` should be 3 (ie `dim+index+1`)
  not 2 (ie currently `dim+index`)

in addition, Pad is incorrectly calculating the starting padding.
If we want to pad out 2 elements to the start, and we have this cycle
of indices of length 6, then we should skip 4 elements, but currently
we skip 2. A more visual representation of what's going on is below:

```
pad_start: 2
data:      [a,b,c,d]
indices:   [0, 1, 2, 3, 2, 1, 0, 1, 2, 3, 2, 1, 0, ..] // zigzag between 0..4
actual:    skip [ c  d| c  b  a  b]
expected:  ~  skip  ~ [ c  b| a  b  c  d]
```

The values between `[` and `|` are padding and the values between
`|` and `]` in the example should match the original data being padded.

* Fix clippy lints.

---------

Co-authored-by: Laurent <laurent.mazare@gmail.com>

candle-examples/examples/beit/main.rs

@@ -16,7 +16,7 @@ use candle_transformers::models::beit;
 /// Loads an image from disk using the image crate, this returns a tensor with shape
 /// (3, 384, 384). Beit special normalization is applied.
 pub fn load_image384_beit_norm<P: AsRef<std::path::Path>>(p: P) -> Result<Tensor> {
-    let img = image::io::Reader::open(p)?
+    let img = image::ImageReader::open(p)?
         .decode()
         .map_err(candle::Error::wrap)?
         .resize_to_fill(384, 384, image::imageops::FilterType::Triangle);

candle-examples/examples/blip/main.rs

@@ -55,7 +55,7 @@ const SEP_TOKEN_ID: u32 = 102;
 /// Loads an image from disk using the image crate, this returns a tensor with shape
 /// (3, 384, 384). OpenAI normalization is applied.
 pub fn load_image<P: AsRef<std::path::Path>>(p: P) -> Result<Tensor> {
-    let img = image::io::Reader::open(p)?
+    let img = image::ImageReader::open(p)?
         .decode()
         .map_err(candle::Error::wrap)?
         .resize_to_fill(384, 384, image::imageops::FilterType::Triangle);

candle-examples/examples/clip/main.rs

@@ -33,7 +33,7 @@ struct Args {
 }
 
 fn load_image<T: AsRef<std::path::Path>>(path: T, image_size: usize) -> anyhow::Result<Tensor> {
-    let img = image::io::Reader::open(path)?.decode()?;
+    let img = image::ImageReader::open(path)?.decode()?;
     let (height, width) = (image_size, image_size);
     let img = img.resize_to_fill(
         width as u32,

candle-examples/examples/eva2/main.rs

@@ -16,7 +16,7 @@ use candle_transformers::models::eva2;
 /// Loads an image from disk using the image crate, this returns a tensor with shape
 /// (3, 448, 448). OpenAI normalization is applied.
 pub fn load_image448_openai_norm<P: AsRef<std::path::Path>>(p: P) -> Result<Tensor> {
-    let img = image::io::Reader::open(p)?
+    let img = image::ImageReader::open(p)?
         .decode()
         .map_err(candle::Error::wrap)?
         .resize_to_fill(448, 448, image::imageops::FilterType::Triangle);

candle-examples/examples/llava/main.rs

@@ -57,7 +57,7 @@ fn load_image<T: AsRef<std::path::Path>>(
     llava_config: &LLaVAConfig,
     dtype: DType,
 ) -> Result<((u32, u32), Tensor)> {
-    let img = image::io::Reader::open(path)?.decode()?;
+    let img = image::ImageReader::open(path)?.decode()?;
     let img_tensor = process_image(&img, processor, llava_config)?;
     Ok(((img.width(), img.height()), img_tensor.to_dtype(dtype)?))
 }

candle-examples/examples/moondream/main.rs

@@ -208,7 +208,7 @@ struct Args {
 /// Loads an image from disk using the image crate, this returns a tensor with shape
 /// (3, 378, 378).
 pub fn load_image<P: AsRef<std::path::Path>>(p: P) -> candle::Result<Tensor> {
-    let img = image::io::Reader::open(p)?
+    let img = image::ImageReader::open(p)?
         .decode()
         .map_err(candle::Error::wrap)?
         .resize_to_fill(378, 378, image::imageops::FilterType::Triangle); // Adjusted to 378x378

candle-examples/examples/segment-anything/main.rs

@@ -139,7 +139,7 @@ pub fn main() -> anyhow::Result<()> {
         let (_one, h, w) = mask.dims3()?;
         let mask = mask.expand((3, h, w))?;
 
-        let mut img = image::io::Reader::open(&args.image)?
+        let mut img = image::ImageReader::open(&args.image)?
             .decode()
             .map_err(candle::Error::wrap)?;
         let mask_pixels = mask.permute((1, 2, 0))?.flatten_all()?.to_vec1::<u8>()?;

candle-examples/examples/stable-diffusion/main.rs

@@ -380,7 +380,7 @@ fn text_embeddings(
 }
 
 fn image_preprocess<T: AsRef<std::path::Path>>(path: T) -> anyhow::Result<Tensor> {
-    let img = image::io::Reader::open(path)?.decode()?;
+    let img = image::ImageReader::open(path)?.decode()?;
     let (height, width) = (img.height() as usize, img.width() as usize);
     let height = height - height % 32;
     let width = width - width % 32;

candle-examples/examples/trocr/image_processor.rs

@@ -145,7 +145,7 @@ impl ViTImageProcessor {
     pub fn load_images(&self, image_path: Vec<&str>) -> Result<Vec<image::DynamicImage>> {
         let mut images: Vec<image::DynamicImage> = Vec::new();
         for path in image_path {
-            let img = image::io::Reader::open(path)?.decode().unwrap();
+            let img = image::ImageReader::open(path)?.decode().unwrap();
             images.push(img);
         }
 

candle-examples/examples/yolo-v3/main.rs

@@ -159,7 +159,7 @@ pub fn main() -> Result<()> {
         let net_width = darknet.width()?;
         let net_height = darknet.height()?;
 
-        let original_image = image::io::Reader::open(&image_name)?
+        let original_image = image::ImageReader::open(&image_name)?
             .decode()
             .map_err(candle::Error::wrap)?;
         let image = {

candle-examples/examples/yolo-v8/main.rs

@@ -390,7 +390,7 @@ pub fn run<T: Task>(args: Args) -> anyhow::Result<()> {
     for image_name in args.images.iter() {
         println!("processing {image_name}");
         let mut image_name = std::path::PathBuf::from(image_name);
-        let original_image = image::io::Reader::open(&image_name)?
+        let original_image = image::ImageReader::open(&image_name)?
             .decode()
             .map_err(candle::Error::wrap)?;
         let (width, height) = {