Add Im2Col support on the gpu side. (#808)

* Add Im2Col support on the gpu side.

* Actually enable.

candle-core/src/cuda_backend.rs

@@ -11,6 +11,8 @@ use cudarc::driver::{
 use half::{bf16, f16};
 use std::sync::{Arc, Mutex};
 
+const USE_IM2COL_CONV2D: bool = true;
+
 /// cudarc related errors
 #[derive(thiserror::Error, Debug)]
 pub enum CudaError {
@@ -1723,8 +1725,37 @@ impl BackendStorage for CudaStorage {
         params: &crate::conv::ParamsConv2D,
     ) -> Result<Self> {
         let device = self.device().clone();
+        if !USE_IM2COL_CONV2D {
             let slice = Conv2D(params).map(&self.slice, l, &kernel.slice, kernel_l, &device)?;
-        Ok(Self { slice, device })
+            return Ok(Self { slice, device });
+        }
+
+        let col = Im2Col {
+            h_k: params.k_h,
+            w_k: params.k_w,
+            stride: params.stride,
+            dilation: params.dilation,
+            padding: params.padding,
+        }
+        .map(&self.slice, &device, l)?;
+        let col = Self { slice: col, device };
+        let h_out = params.out_h();
+        let w_out = params.out_w();
+        let b = params.b_size;
+        let n = params.c_out;
+        let k = params.k_h * params.k_w * params.c_in;
+        let m = h_out * w_out;
+        let col_l = Layout::contiguous((b, m, k));
+        let kernel_l = Layout::contiguous_with_offset((1, n, k), kernel_l.start_offset())
+            .transpose(1, 2)?
+            .broadcast_as((b, k, n))?;
+        let res = col.matmul(kernel, (b, m, n, k), &col_l, &kernel_l)?;
+        let res_l = Layout::contiguous((b, h_out, w_out, n))
+            .transpose(1, 2)?
+            .transpose(1, 3)?;
+        let mut res_t = self.device().zeros_impl(res_l.shape(), res.dtype())?;
+        res.copy_strided_src(&mut res_t, 0, &res_l)?;
+        Ok(res_t)
     }
 
     #[cfg(feature = "cudnn")]