github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-22 21:38:10 -05:00
Code Issues Packages Projects Releases Wiki Activity

remove the expand

Browse Source
This commit is contained in:
George Hotz
2022-06-16 09:54:56 -07:00
parent bcfbb4c81b
commit 38d6cfec2a
3 changed files with 10 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tinygrad/llops/ops_gpu.py
View File

@@ -93,7 +93,7 @@ def reduce_op(op, inp, ret):
# this takes a ret index to an inp index, indexing 0 on the reduced strides
view = View(ret.shape, strides_for_shape(inp.shape))
# combined adjacent reduce axis
# generate loops with combined adjacent reduce axis
acc = 1
loop_start, loop_end = [], []
for shp,stride in st.views[-1].shape_strides[::-1]:

19
tinygrad/mlops.py
View File

@@ -192,19 +192,14 @@ class Conv2D(Function):
# compute derivative of weights using ProcessingOps.CONV
# TODO: there has to be a way to do this without the expand/reduce for at least matmul
# since it's ctx.op.matmul(input, grad_output, ctx.buffer(weight.shape), transpose_a=True)
xdw = ctx.movement_op(MovementOps.RESHAPE, x, (1, C.bs * C.groups * C.cin, C.iy, C.ix))
grad_output_dw = ctx.movement_op(MovementOps.RESHAPE, grad_output, (C.bs * C.groups, 1, C.rcout, C.oy, C.ox))
# this expand is slow
grad_output_dw = ctx.movement_op(MovementOps.EXPAND, grad_output_dw, (C.bs * C.groups, C.cin, C.rcout, C.oy, C.ox))
grad_output_dw = ctx.movement_op(MovementOps.RESHAPE, grad_output_dw, (C.bs * C.groups * C.cin * C.rcout, 1, C.oy, C.ox))
# padding is the same, stride and dilation are flipped
Cdw = get_conv_args(xdw.shape, grad_output_dw.shape, padding=(C.py, C.px), stride=(C.dy, C.dx), dilation=(C.ys, C.xs), groups=C.bs*C.groups*C.cin)
xdw = ctx.movement_op(MovementOps.RESHAPE, x, (C.bs*C.groups, C.cin, C.iy, C.ix))
xdw = ctx.movement_op(MovementOps.PERMUTE, xdw, (1,0,2,3))
grad_output_dw = ctx.movement_op(MovementOps.RESHAPE, grad_output, (C.bs * C.groups * C.rcout, 1, C.oy, C.ox))
Cdw = get_conv_args(xdw.shape, grad_output_dw.shape, padding=(C.py, C.px), stride=(C.dy, C.dx), dilation=(C.ys, C.xs), groups=C.bs*C.groups)
grad_weight = ctx.processing_op(ProcessingOps.CONV, xdw, grad_output_dw, (Cdw.bs, Cdw.cout, Cdw.oy, Cdw.ox), Cdw)
grad_weight = ctx.movement_op(MovementOps.RESHAPE, grad_weight, (C.bs, C.groups, C.cin, C.rcout, Cdw.oy, Cdw.ox))
# sum across the batch dimension
grad_weight = ctx.movement_op(MovementOps.RESHAPE, grad_weight, (C.cin, C.bs, C.groups, C.rcout, Cdw.oy, Cdw.ox))
grad_weight = ctx.movement_op(MovementOps.PERMUTE, grad_weight, (1,2,3,0,4,5))
grad_weight = ctx.reduce_op(ReduceOps.SUM, grad_weight, (1, *grad_weight.shape[1:]))
# flip channels out and in
grad_weight = ctx.movement_op(MovementOps.PERMUTE, grad_weight, (0,1,3,2,4,5))
grad_weight = ctx.movement_op(MovementOps.RESHAPE, grad_weight, (C.groups*C.rcout, C.cin, Cdw.oy, Cdw.ox))
dw = ctx.movement_op(MovementOps.SLICE, grad_weight, ((0, grad_weight.shape[0]), (0, grad_weight.shape[1]), (0, w.shape[2]), (0, w.shape[3])))
return dx, dw
return dx, dw

4
tinygrad/ops.py
View File

@@ -48,13 +48,13 @@ class Ops:
log_op(op, ret, [x])
return ret
def reduce_op(ctx, op:BinaryOps, x, new_shape):
def reduce_op(ctx, op:ReduceOps, x, new_shape):
ret = ctx.buffer(new_shape)
ctx.op.reduce_op(op, x, ret)
log_op(op, ret, [x])
return ret
def binary_op(ctx, op:ReduceOps, x, y):
def binary_op(ctx, op:BinaryOps, x, y):
assert x.shape == y.shape
ret = ctx.buffer(x.shape)
ctx.op.binary_op(op, x, y, ret)