Moving to gemm and adding matmul backprop.

- Tentative `T` operator.

Cargo.toml

@@ -16,11 +16,11 @@ members = [
 ]
 
 [dependencies]
-ggblas = "0.1.0"
 safetensors = "0.3.1"
 thiserror = "1"
 cudarc = { version = "0.9.9", optional = true }
 candle-kernels = { path = "kernels", optional = true }
+gemm = "0.15.4"
 
 [dev-dependencies]
 anyhow = "1"

src/cpu_backend.rs

@@ -1,6 +1,6 @@
 use crate::storage::{BinaryOp, UnaryOp};
 use crate::{DType, Error, Result, Shape, StridedIndex};
-use ggblas::batched_sgemm;
+use gemm::{gemm, Parallelism};
 
 // TODO: Think about whether we would be better off with a dtype and
 // a buffer as an owned slice of bytes.
@@ -113,28 +113,83 @@ impl CpuStorage {
         lhs_stride: &[usize],
         rhs_stride: &[usize],
     ) -> Result<Self> {
-        println!("rhs {rhs:?}");
-        println!("lhs_stride {lhs_stride:?}");
-        println!("rhs_stride {rhs_stride:?}");
-        // todo!("matmul");
         let a_skip: usize = m * k;
         let b_skip: usize = n * k;
         let c_skip: usize = m * n;
 
-        let mut c = Self::F32(vec![0.0; b * m * n]);
+        let rank = lhs_stride.len();
+        let lhs_cs = lhs_stride[rank - 1];
+        let lhs_rs = lhs_stride[rank - 2];
 
-        batched_sgemm(
+        let rhs_cs = rhs_stride[rank - 1];
-            self.as_slice()?,
+        let rhs_rs = rhs_stride[rank - 2];
-            a_skip,
+
-            rhs.as_slice()?,
+        if lhs_stride.len() > 2 {
-            b_skip,
+            let lhs_batch_stride = &lhs_stride[..rank - 2];
-            c.as_mut_slice()?,
+            let rhs_batch_stride = &rhs_stride[..rank - 2];
-            c_skip,
+
-            m,
+            if lhs_batch_stride != &[a_skip] || rhs_batch_stride != &[b_skip] {
-            n,
+                // Temporary error before we support abitrary striding.
-            k,
+                return Err(Error::UnexpectedStriding);
-            b,
+            }
-        );
+        }
+
+        let mut dst = vec![0.0; b * m * n];
+
+        let dst_shape: Shape = (m, n).into();
+        let dst_strides = dst_shape.stride_contiguous();
+        let dst_rs = dst_strides[0];
+        let dst_cs = dst_strides[1];
+
+        for step in 0..b {
+            let lhs_p = &self.as_slice::<f32>()?[step * a_skip..];
+            let rhs_p = &rhs.as_slice::<f32>()?[step * b_skip..];
+            let dst_p = &mut dst[step * c_skip..];
+            unsafe {
+                gemm(
+                    // m: usize,
+                    m,
+                    // n: usize,
+                    n,
+                    // k: usize,
+                    k,
+                    // dst: *mut T,
+                    dst_p.as_mut_ptr(),
+                    // dst_cs: isize,
+                    dst_cs as isize,
+                    // dst_rs: isize,
+                    dst_rs as isize,
+                    // read_dst: bool,
+                    false,
+                    // lhs: *const T,
+                    lhs_p.as_ptr(),
+                    // lhs_cs: isize,
+                    lhs_cs as isize,
+                    // lhs_rs: isize,
+                    lhs_rs as isize,
+                    // rhs: *const T,
+                    rhs_p.as_ptr(),
+                    // rhs_cs: isize,
+                    rhs_cs as isize,
+                    // rhs_rs: isize,
+                    rhs_rs as isize,
+                    // alpha: T,
+                    1.0,
+                    // beta: T,
+                    1.0,
+                    // conj_dst: bool,
+                    false,
+                    // conj_lhs: bool,
+                    false,
+                    // conj_rhs: bool,
+                    true,
+                    // parallelism: Parallelism
+                    Parallelism::None,
+                )
+            }
+        }
+
+        let c = Self::F32(dst);
         Ok(c)
     }
 
@@ -175,31 +230,31 @@ mod tests {
     #[test]
     fn simple_matmul() -> Result<()> {
         let data = vec![1.0f32, 2.0, 3.0, 4.0];
-        let a = Tensor::from_slice(&data, (2, 2), Device::Cpu)?;
+        let a = Tensor::from_slice(&data, (2, 2), &Device::Cpu)?;
         let data = vec![1.0f32, 2.0, 3.0, 4.0];
-        let b = Tensor::from_slice(&data, (2, 2), Device::Cpu)?;
+        let b = Tensor::from_slice(&data, (2, 2), &Device::Cpu)?;
 
         let c = a.matmul(&b)?;
         assert_eq!(c.to_vec2::<f32>()?, &[&[7.0f32, 10.0], &[15.0, 22.0]]);
 
         let data = vec![1.0f32, 2.0];
-        let a = Tensor::from_slice(&data, (2, 1), Device::Cpu)?;
+        let a = Tensor::from_slice(&data, (2, 1), &Device::Cpu)?;
         let data = vec![3.0f32, 4.0];
-        let b = Tensor::from_slice(&data, (1, 2), Device::Cpu)?;
+        let b = Tensor::from_slice(&data, (1, 2), &Device::Cpu)?;
         let c = a.matmul(&b)?;
         assert_eq!(c.to_vec2::<f32>()?, &[&[3.0, 4.0], &[6.0, 8.0]]);
 
         let data: Vec<_> = (0..6).map(|i| i as f32).collect();
-        let a = Tensor::from_slice(&data, (2, 3), Device::Cpu)?;
+        let a = Tensor::from_slice(&data, (2, 3), &Device::Cpu)?;
         let data: Vec<_> = (0..6).map(|i| (i + 2) as f32).collect();
-        let b = Tensor::from_slice(&data, (3, 2), Device::Cpu)?;
+        let b = Tensor::from_slice(&data, (3, 2), &Device::Cpu)?;
         let c = a.matmul(&b)?;
         assert_eq!(c.to_vec2::<f32>()?, &[&[16., 19.], &[52., 64.]]);
 
         let data: Vec<_> = (0..12).map(|i| i as f32).collect();
-        let a = Tensor::from_slice(&data, (2, 2, 3), Device::Cpu)?;
+        let a = Tensor::from_slice(&data, (2, 2, 3), &Device::Cpu)?;
         let data: Vec<_> = (0..12).map(|i| (i + 2) as f32).collect();
-        let b = Tensor::from_slice(&data, (2, 3, 2), Device::Cpu)?;
+        let b = Tensor::from_slice(&data, (2, 3, 2), &Device::Cpu)?;
         let c = a.matmul(&b)?;
         assert_eq!(
             c.to_vec3::<f32>()?,

src/error.rs

@@ -40,6 +40,9 @@ pub enum Error {
         shape: Shape,
     },
 
+    #[error("temporary error where matmul doesn't support arbitrary striding")]
+    UnexpectedStriding,
+
     #[error(transparent)]
     Cuda(#[from] crate::CudaError),
 }

src/tensor.rs

@@ -147,10 +147,11 @@ impl Tensor {
 
     pub fn new_impl<A: crate::device::NdArray>(
         array: A,
+        shape: Shape,
         device: &Device,
         is_variable: bool,
     ) -> Result<Self> {
-        let shape = array.shape()?;
+        // let shape = array.shape()?;
         let storage = device.storage(array)?;
         let stride = shape.stride_contiguous();
         let tensor_ = Tensor_ {
@@ -165,31 +166,29 @@ impl Tensor {
     }
 
     pub fn new<A: crate::device::NdArray>(array: A, device: &Device) -> Result<Self> {
-        Self::new_impl(array, device, false)
+        let shape = array.shape()?.clone();
+        Self::new_impl(array, shape, device, false)
     }
 
     pub fn var<A: crate::device::NdArray>(array: A, device: &Device) -> Result<Self> {
-        Self::new_impl(array, device, true)
+        let shape = array.shape()?.clone();
+        Self::new_impl(array, shape, device, true)
     }
 
     pub fn from_slice<S: Into<Shape>, D: crate::WithDType>(
-        a: &[D],
+        array: &[D],
         shape: S,
-        device: Device,
+        device: &Device,
     ) -> Result<Self> {
-        let shape = shape.into();
+        Self::new_impl(array, shape.into(), device, false)
-        let storage = device.storage(a)?;
+    }
-        let stride = shape.stride_contiguous();
+
-        let is_variable = false;
+    pub fn var_from_slice<S: Into<Shape>, D: crate::WithDType>(
-        let tensor_ = Tensor_ {
+        array: &[D],
-            id: TensorId::new(),
+        shape: S,
-            storage,
+        device: &Device,
-            shape,
+    ) -> Result<Self> {
-            stride,
+        Self::new_impl(array, shape.into(), device, true)
-            op: None,
-            is_variable,
-        };
-        Ok(Self(Arc::new(tensor_)))
     }
 
     pub(crate) fn same_shape_binary_op(&self, rhs: &Self, op: &'static str) -> Result<&Shape> {
@@ -260,6 +259,7 @@ impl Tensor {
 
         let dim = a_dims.len();
 
+        // TODO
         // if dim < 2 {
         //     return Err(SmeltError::InsufficientRank { minimum_rank: 2 });
         // }
@@ -309,6 +309,13 @@ impl Tensor {
         crate::StridedIndex::new(self.dims(), self.stride())
     }
 
+    pub fn as_slice<S: crate::WithDType>(&self) -> Result<&[S]> {
+        match &self.storage {
+            Storage::Cpu(cpu_storage) => S::cpu_storage_as_slice(cpu_storage),
+            Storage::Cuda { .. } => todo!(),
+        }
+    }
+
     pub fn to_vec1<S: crate::WithDType>(&self) -> Result<Vec<S>> {
         if self.rank() != 1 {
             return Err(Error::UnexpectedNumberOfDims {
@@ -404,6 +411,31 @@ impl Tensor {
         self.id
     }
 
+    pub fn t(&self) -> Result<Tensor> {
+        let mut stride = self.stride().to_vec();
+        let mut shape = self.shape().clone();
+        let n = stride.len();
+        if n < 2 {
+            return Err(Error::UnexpectedNumberOfDims {
+                expected: 2,
+                got: n,
+                shape: self.shape().clone(),
+            });
+        }
+        (shape.0[n - 2], shape.0[n - 1]) = (shape.0[n - 1], shape.0[n - 2]);
+        (stride[n - 2], stride[n - 1]) = (stride[n - 1], stride[n - 2]);
+        let tensor_ = Tensor_ {
+            id: TensorId::new(),
+            storage: self.storage.clone(),
+            shape,
+            stride,
+            // TODO The op should have a backward
+            op: None,
+            is_variable: false,
+        };
+        Ok(Tensor(Arc::new(tensor_)))
+    }
+
     pub fn is_contiguous(&self) -> bool {
         self.shape.is_contiguous(&self.stride)
     }
@@ -514,37 +546,17 @@ impl Tensor {
                         let rhs_sum_grad = grads.or_insert(rhs)?;
                         *rhs_sum_grad = rhs_sum_grad.add(&rhs_grad)?;
                     }
-                    Op::Matmul(_lhs, _rhs) => {
+                    Op::Matmul(lhs, rhs) => {
-                        // let (m, k) = lhs.shape;
+                        // Skipping checks, the op went ok, we can skip
-                        // let n = rhs.shape.1;
+                        // the matmul size checks for now.
-                        // let strides = (m, n).strides();
-                        // Self::matmul(
-                        //     (m, n, k),
-                        //     true,
-                        //     grad_out.as_ptr(),
-                        //     strides,
-                        //     rhs.data.as_ptr(),
-                        //     [rhs.strides[1], rhs.strides[0]],
-                        //     grad_lhs.as_mut_ptr(),
-                        //     lhs.strides,
-                        // );
-                        // Self::matmul(
-                        //     (k, m, n),
-                        //     true,
-                        //     lhs.data.as_ptr(),
-                        //     [lhs.strides[1], lhs.strides[0]],
-                        //     grad_out.as_ptr(),
-                        //     strides,
-                        //     grad_rhs.as_mut_ptr(),
-                        //     rhs.strides,
-                        // );
 
-                        // let lhs_grad = grad.matmul(rhs)?;
+                        let lhs_grad = grad.matmul(&rhs.t()?)?;
-                        // let lhs_sum_grad = grads.entry(lhs.id).or_insert_with(|| lhs.zeros_like());
+                        let lhs_sum_grad = grads.or_insert(lhs)?;
-                        // *lhs_sum_grad = lhs_sum_grad.add(&lhs_grad)?;
+                        *lhs_sum_grad = lhs_sum_grad.add(&lhs_grad)?;
-                        // let rhs_grad = grad.mul(lhs)?.div(&rhs.sqr()?)?;
+
-                        // let rhs_sum_grad = grads.entry(rhs.id).or_insert_with(|| rhs.zeros_like());
+                        let rhs_grad = lhs.t()?.matmul(&grad)?;
-                        // *rhs_sum_grad = rhs_sum_grad.add(&rhs_grad)?;
+                        let rhs_sum_grad = grads.or_insert(rhs)?;
+                        *rhs_sum_grad = rhs_sum_grad.add(&rhs_grad)?;
                     }
                     Op::Affine { arg, mul, .. } => {
                         let arg_grad = grad.affine(*mul, 0.)?;

tests/grad_tests.rs

@@ -1,5 +1,5 @@
 use anyhow::{Context, Result};
-use candle::{Device, Tensor};
+use candle::{Device, Shape, Tensor};
 
 #[test]
 fn simple_grad() -> Result<()> {
@@ -14,3 +14,27 @@ fn simple_grad() -> Result<()> {
     assert_eq!(grad_x.to_vec1::<f32>()?, [11., 7., 13.]);
     Ok(())
 }
+
+#[test]
+fn matmul_grad() -> Result<()> {
+    let data: Vec<_> = (0..12).map(|i| i as f32).collect();
+    let x = Tensor::var_from_slice(&data, (2, 2, 3), &Device::Cpu)?;
+    let data: Vec<_> = (0..12).map(|i| i as f32).collect();
+    let y = Tensor::var_from_slice(&data, (2, 3, 2), &Device::Cpu)?;
+
+    let c = x.matmul(&y)?;
+    let grads = c.backward()?;
+    let grad_x = grads.get(&x).context("no grad for x")?;
+    let grad_y = grads.get(&y).context("no grad for y")?;
+    assert_eq!(grad_x.shape(), &Shape::from((2, 2, 3)));
+    assert_eq!(grad_y.shape(), &Shape::from((2, 3, 2)));
+    assert_eq!(
+        grad_x.as_slice::<f32>()?,
+        &[1., 5., 9., 1., 5., 9., 13., 17., 21., 13., 17., 21.]
+    );
+    assert_eq!(
+        grad_y.as_slice::<f32>()?,
+        &[3., 3., 5., 5., 7., 7., 15., 15., 17., 17., 19., 19.]
+    );
+    Ok(())
+}