Add some fast Metal MLX SDPA kernels (#2584)

* Add some fast Metal MLX SDPA kernels (#32)

* Sketch the sdpa kernel

* Add full sdpa kernel,

* Add test

* Add vectorized kernel for decoding

* Update tests

* Add some docs

* Fix sdpa_vector names

* Add softcapping for vectorized sdpa

* Add softcapping for full sdpa

* Add support for head dim 32, 96, 256

* Add support for head dim 32, 96, 256

* Update docs

* Add update notice

* Clippy and format

* Conditional compilation for bf16

* Use it in quantized llama

* Some review comments

* Use set_params!

* Remove unused

* Remove feature

* Fix metal sdpa for v stride

* Remove comma

* Add the dim method to layout and shape.

---------

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

candle-core/src/layout.rs

@@ -35,6 +35,12 @@ impl Layout {
         self.shape.dims()
     }
 
+    /// The dimension size for a specified dimension index.
+    pub fn dim<D: crate::shape::Dim>(&self, dim: D) -> Result<usize> {
+        let dim = dim.to_index(&self.shape, "dim")?;
+        Ok(self.dims()[dim])
+    }
+
     pub fn shape(&self) -> &Shape {
         &self.shape
     }