Note
Click here to download the full example code
Layer NormalizationΒΆ
Out:
layer-norm-backward:
N Triton Torch
0 1024.0 307.200008 98.303995
1 1536.0 347.773587 134.540150
2 2048.0 420.102553 161.154101
3 2560.0 455.111129 181.238943
4 3072.0 511.999982 192.501302
5 3584.0 551.384634 208.271186
6 4096.0 564.965515 220.412561
7 4608.0 504.986315 233.316456
8 5120.0 527.381977 242.366855
9 5632.0 542.843364 243.545956
10 6144.0 546.133354 248.661056
11 6656.0 532.479975 256.410903
12 7168.0 507.469040 260.260201
13 7680.0 482.513091 262.938666
14 8192.0 463.698115 265.327937
15 8704.0 416.958106 267.815384
16 9216.0 431.157889 271.391419
17 9728.0 438.857162 280.615388
18 10240.0 448.467168 286.433562
19 10752.0 428.651173 246.699797
20 11264.0 427.071098 245.760001
21 11776.0 423.089806 249.667843
22 12288.0 419.504980 254.673582
23 12800.0 413.458944 253.674644
24 13312.0 411.711355 252.161013
25 13824.0 405.594132 257.190689
26 14336.0 394.568805 254.485198
27 14848.0 386.080180 257.479779
28 15360.0 374.634130 257.970599
29 15872.0 368.402336 262.166551
import torch
import triton
import triton.language as tl
try:
# This is https://github.com/NVIDIA/apex, NOT the apex on PyPi, so it
# should not be added to extras_require in setup.py.
import apex
HAS_APEX = True
except ModuleNotFoundError:
HAS_APEX = False
# Forward Pass
@triton.jit
def _layer_norm_fwd_fused(X, Y, W, B, M, V, stride, N, eps,
BLOCK_SIZE: tl.constexpr):
# position of elements processed by this program
row = tl.program_id(0)
cols = tl.arange(0, BLOCK_SIZE)
mask = cols < N
# offset data pointers to start at the row of interest
X += row * stride
Y += row * stride
# load data and cast to float32
x = tl.load(X + cols, mask=mask, other=0).to(tl.float32)
# compute mean
mean = tl.sum(x, axis=0) / N
# compute std
xmean = tl.where(mask, x - mean, 0.)
var = tl.sum(xmean * xmean, axis=0) / N
rstd = 1 / tl.sqrt(var + eps)
xhat = xmean * rstd
# write-back mean/rstd
tl.store(M + row, mean)
tl.store(V + row, rstd)
# multiply by weight and add bias
w = tl.load(W + cols, mask=mask)
b = tl.load(B + cols, mask=mask)
y = xhat * w + b
# write-back
tl.store(Y + cols, y, mask=mask)
# Backward pass (DX + partial DW + partial DB)
@triton.jit
def _layer_norm_bwd_dx_fused(DX, DY, DW, DB, X, W, B, M, V, Lock, stride, N, eps,
GROUP_SIZE_M: tl.constexpr, BLOCK_SIZE_N: tl.constexpr):
# position of elements processed by this program
row = tl.program_id(0)
cols = tl.arange(0, BLOCK_SIZE_N)
mask = cols < N
# offset data pointers to start at the row of interest
X += row * stride
DY += row * stride
DX += row * stride
# offset locks and weight/bias gradient pointer
# each kernel instance accumulates partial sums for
# DW and DB into one of GROUP_SIZE_M independent buffers
# these buffers stay in the L2, which allow this kernel
# to be fast
lock_id = row % GROUP_SIZE_M
Lock += lock_id
Count = Lock + GROUP_SIZE_M
DW = DW + lock_id * N + cols
DB = DB + lock_id * N + cols
# load data to SRAM
x = tl.load(X + cols, mask=mask, other=0).to(tl.float32)
dy = tl.load(DY + cols, mask=mask, other=0).to(tl.float32)
w = tl.load(W + cols, mask=mask).to(tl.float32)
mean = tl.load(M + row)
rstd = tl.load(V + row)
# compute dx
xhat = (x - mean) * rstd
wdy = w * dy
xhat = tl.where(mask, xhat, 0.)
wdy = tl.where(mask, wdy, 0.)
mean1 = tl.sum(xhat * wdy, axis=0) / N
mean2 = tl.sum(wdy, axis=0) / N
dx = (wdy - (xhat * mean1 + mean2)) * rstd
# write-back dx
tl.store(DX + cols, dx, mask=mask)
# accumulate partial sums for dw/db
partial_dw = (dy * xhat).to(w.dtype)
partial_db = (dy).to(w.dtype)
while tl.atomic_cas(Lock, 0, 1) == 1:
pass
count = tl.load(Count)
# first store doesn't accumulate
if count == 0:
tl.atomic_xchg(Count, 1)
else:
partial_dw += tl.load(DW, mask=mask)
partial_db += tl.load(DB, mask=mask)
tl.store(DW, partial_dw, mask=mask)
tl.store(DB, partial_db, mask=mask)
# release lock
tl.atomic_xchg(Lock, 0)
# Backward pass (total DW + total DB)
@triton.jit
def _layer_norm_bwd_dwdb(DW, DB, FINAL_DW, FINAL_DB, M, N,
BLOCK_SIZE_M: tl.constexpr, BLOCK_SIZE_N: tl.constexpr):
pid = tl.program_id(0)
cols = pid * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N)
dw = tl.zeros((BLOCK_SIZE_M, BLOCK_SIZE_N), dtype=tl.float32)
db = tl.zeros((BLOCK_SIZE_M, BLOCK_SIZE_N), dtype=tl.float32)
for i in range(0, M, BLOCK_SIZE_M):
rows = i + tl.arange(0, BLOCK_SIZE_M)
mask = (rows[:, None] < M) & (cols[None, :] < N)
offs = rows[:, None] * N + cols[None, :]
dw += tl.load(DW + offs, mask=mask, other=0.)
db += tl.load(DB + offs, mask=mask, other=0.)
sum_dw = tl.sum(dw, axis=0)
sum_db = tl.sum(db, axis=0)
tl.store(FINAL_DW + cols, sum_dw, mask=cols < N)
tl.store(FINAL_DB + cols, sum_db, mask=cols < N)
class LayerNorm(torch.autograd.Function):
@staticmethod
def forward(ctx, x, normalized_shape, weight, bias, eps):
# allocate output
y = torch.empty_like(x)
# reshape input data into 2D tensor
x_arg = x.reshape(-1, x.shape[-1])
M, N = x_arg.shape
mean = torch.empty((M, ), dtype=torch.float32, device='cuda')
rstd = torch.empty((M, ), dtype=torch.float32, device='cuda')
# Less than 64KB per feature: enqueue fused kernel
MAX_FUSED_SIZE = 65536 // x.element_size()
BLOCK_SIZE = min(MAX_FUSED_SIZE, triton.next_power_of_2(N))
if N > BLOCK_SIZE:
raise RuntimeError("This layer norm doesn't support feature dim >= 64KB.")
# heuristics for number of warps
num_warps = min(max(BLOCK_SIZE // 256, 1), 8)
# enqueue kernel
_layer_norm_fwd_fused[(M,)](x_arg, y, weight, bias, mean, rstd,
x_arg.stride(0), N, eps,
BLOCK_SIZE=BLOCK_SIZE, num_warps=num_warps)
ctx.save_for_backward(x, weight, bias, mean, rstd)
ctx.BLOCK_SIZE = BLOCK_SIZE
ctx.num_warps = num_warps
ctx.eps = eps
return y
@staticmethod
def backward(ctx, dy):
x, w, b, m, v = ctx.saved_tensors
# heuristics for amount of parallel reduction stream for DG/DB
N = w.shape[0]
GROUP_SIZE_M = 64
if N <= 8192: GROUP_SIZE_M = 96
if N <= 4096: GROUP_SIZE_M = 128
if N <= 1024: GROUP_SIZE_M = 256
# allocate output
locks = torch.zeros(2 * GROUP_SIZE_M, dtype=torch.int32, device='cuda')
_dw = torch.empty((GROUP_SIZE_M, w.shape[0]), dtype=x.dtype, device=w.device)
_db = torch.empty((GROUP_SIZE_M, w.shape[0]), dtype=x.dtype, device=w.device)
dw = torch.empty((w.shape[0],), dtype=w.dtype, device=w.device)
db = torch.empty((w.shape[0],), dtype=w.dtype, device=w.device)
dx = torch.empty_like(dy)
# enqueue kernel using forward pass heuristics
# also compute partial sums for DW and DB
x_arg = x.reshape(-1, x.shape[-1])
M, N = x_arg.shape
_layer_norm_bwd_dx_fused[(M,)](dx, dy, _dw, _db, x, w, b, m, v, locks,
x_arg.stride(0), N, ctx.eps,
BLOCK_SIZE_N=ctx.BLOCK_SIZE,
GROUP_SIZE_M=GROUP_SIZE_M,
num_warps=ctx.num_warps)
grid = lambda meta: [triton.cdiv(N, meta['BLOCK_SIZE_N'])]
# accumulate partial sums in separate kernel
_layer_norm_bwd_dwdb[grid](_dw, _db, dw, db, GROUP_SIZE_M, N,
BLOCK_SIZE_M=32,
BLOCK_SIZE_N=128)
return dx, None, dw, db, None
layer_norm = LayerNorm.apply
def test_layer_norm(M, N, dtype, eps=1e-5, device='cuda'):
# create data
x_shape = (M, N)
w_shape = (x_shape[-1], )
weight = torch.rand(w_shape, dtype=dtype, device='cuda', requires_grad=True)
bias = torch.rand(w_shape, dtype=dtype, device='cuda', requires_grad=True)
x = -2.3 + 0.5 * torch.randn(x_shape, dtype=dtype, device='cuda')
dy = .1 * torch.randn_like(x)
x.requires_grad_(True)
# forward pass
y_tri = layer_norm(x, w_shape, weight, bias, eps)
y_ref = torch.nn.functional.layer_norm(x, w_shape, weight, bias, eps).to(dtype)
# backward pass (triton)
y_tri.backward(dy, retain_graph=True)
dx_tri, dw_tri, db_tri = [_.grad.clone() for _ in [x, weight, bias]]
x.grad, weight.grad, bias.grad = None, None, None
# backward pass (torch)
y_ref.backward(dy, retain_graph=True)
dx_ref, dw_ref, db_ref = [_.grad.clone() for _ in [x, weight, bias]]
# compare
triton.testing.assert_almost_equal(y_tri, y_ref)
triton.testing.assert_almost_equal(dx_tri, dx_ref)
triton.testing.assert_almost_equal(db_tri, db_ref, decimal=1)
triton.testing.assert_almost_equal(dw_tri, dw_ref, decimal=1)
@triton.testing.perf_report(
triton.testing.Benchmark(
x_names=['N'],
x_vals=[512 * i for i in range(2, 32)],
line_arg='provider',
line_vals=['triton', 'torch'] + (['apex'] if HAS_APEX else []),
line_names=['Triton', 'Torch'] + (['Apex'] if HAS_APEX else []),
styles=[('blue', '-'), ('green', '-'), ('orange', '-')],
ylabel='GB/s',
plot_name='layer-norm-backward',
args={'M': 4096, 'dtype': torch.float16, 'mode': 'backward'}
)
)
def bench_layer_norm(M, N, dtype, provider, mode='backward', eps=1e-5, device='cuda'):
# create data
x_shape = (M, N)
w_shape = (x_shape[-1], )
weight = torch.rand(w_shape, dtype=dtype, device='cuda', requires_grad=True)
bias = torch.rand(w_shape, dtype=dtype, device='cuda', requires_grad=True)
x = -2.3 + 0.5 * torch.randn(x_shape, dtype=dtype, device='cuda')
dy = .1 * torch.randn_like(x)
x.requires_grad_(True)
# utility functions
if provider == 'triton':
y_fwd = lambda: layer_norm(x, w_shape, weight, bias, eps)
if provider == 'torch':
y_fwd = lambda: torch.nn.functional.layer_norm(x, w_shape, weight, bias, eps)
if provider == 'apex':
apex_layer_norm = apex.normalization.FusedLayerNorm(w_shape).to(x.device).to(x.dtype)
y_fwd = lambda: apex_layer_norm(x)
# forward pass
if mode == 'forward':
gbps = lambda ms: 2 * x.numel() * x.element_size() / ms * 1e-6
ms, min_ms, max_ms = triton.testing.do_bench(y_fwd, rep=500)
# backward pass
if mode == 'backward':
gbps = lambda ms: 3 * x.numel() * x.element_size() / ms * 1e-6
y = y_fwd()
ms, min_ms, max_ms = triton.testing.do_bench(lambda: y.backward(dy, retain_graph=True),
grad_to_none=[x], rep=500)
return gbps(ms), gbps(max_ms), gbps(min_ms)
bench_layer_norm.run(save_path='.', print_data=True)
Total running time of the script: ( 1 minutes 22.248 seconds)