conv stride support

This commit is contained in:
George Hotz 2020-10-26 08:54:43 -07:00
parent 2a55d7402b
commit 1654008c1f
2 changed files with 12 additions and 6 deletions

View File

@ -40,7 +40,6 @@ class TestOps(unittest.TestCase):
lambda x,w: torch.nn.functional.conv2d(x,w).relu(), lambda x,w: torch.nn.functional.conv2d(x,w).relu(),
lambda x,w: Tensor.conv2d(x,w).relu(), atol=2e-5, grad_atol=2e-6) lambda x,w: Tensor.conv2d(x,w).relu(), atol=2e-5, grad_atol=2e-6)
@unittest.skip("please write stride support")
def test_strided_conv2d(self): def test_strided_conv2d(self):
bs = 4 bs = 4
cin = 3 cin = 3

View File

@ -102,16 +102,21 @@ register('logsoftmax', LogSoftmax)
class Conv2D(Function): class Conv2D(Function):
@staticmethod @staticmethod
def forward(ctx, x, w): def forward(ctx, x, w, stride=1):
if type(ctx.stride) == int:
ctx.stride = (ctx.stride, ctx.stride)
cout,cin,H,W = w.shape cout,cin,H,W = w.shape
tw = w.reshape(cout, -1).T tw = w.reshape(cout, -1).T
bs,oy,ox = x.shape[0], x.shape[2]-(H-1), x.shape[3]-(W-1) ys,xs = ctx.stride
bs,oy,ox = x.shape[0], (x.shape[2]-(H-ys))//ys, (x.shape[3]-(W-xs))//xs
ctx.save_for_backward(x, w) ctx.save_for_backward(x, w)
ret = np.zeros((bs, cout, oy, ox), dtype=w.dtype) ret = np.zeros((bs, cout, oy, ox), dtype=w.dtype)
for Y in range(oy): for Y in range(oy):
for X in range(ox): for X in range(ox):
tx = x[:, :, Y:Y+H, X:X+W].reshape(bs, -1) iY,iX = Y*ys, X*xs
tx = x[:, :, iY:iY+H, iX:iX+W].reshape(bs, -1)
ret[:, :, Y, X] = tx.dot(tw) ret[:, :, Y, X] = tx.dot(tw)
return ret return ret
@ -121,14 +126,16 @@ class Conv2D(Function):
x, w = ctx.saved_tensors x, w = ctx.saved_tensors
cout,cin,H,W = w.shape cout,cin,H,W = w.shape
tw = w.reshape(cout, -1) tw = w.reshape(cout, -1)
ys,xs = ctx.stride
dx, dw = np.zeros_like(x), np.zeros_like(w) dx, dw = np.zeros_like(x), np.zeros_like(w)
for Y in range(grad_output.shape[2]): for Y in range(grad_output.shape[2]):
for X in range(grad_output.shape[3]): for X in range(grad_output.shape[3]):
iY,iX = Y*ys, X*xs
gg = grad_output[:, :, Y, X] gg = grad_output[:, :, Y, X]
tx = x[:, :, Y:Y+H, X:X+W].reshape(x.shape[0], -1) tx = x[:, :, iY:iY+H, iX:iX+W].reshape(x.shape[0], -1)
dw += gg.T.dot(tx).reshape(dw.shape) dw += gg.T.dot(tx).reshape(dw.shape)
dx[:, :, Y:Y+H, X:X+W] += gg.dot(tw).reshape(dx.shape[0], dx.shape[1], H, W) dx[:, :, iY:iY+H, iX:iX+W] += gg.dot(tw).reshape(dx.shape[0], dx.shape[1], H, W)
return dx, dw return dx, dw
register('conv2d', Conv2D) register('conv2d', Conv2D)