example_linear_attn_bwd.py 7.58 KB
Newer Older
1
2
3
4
5
6
7
8
9
import torch
import tilelang as tl
import tilelang.language as T
from tilelang.profiler import do_bench

import argparse
from fla.ops.linear_attn import fused_chunk_linear_attn  # We compare with FLA


10
11
12
13
14
15
@tl.jit(
    out_idx=[4, 5, 6],
    pass_configs={
        "tl.disable_tma_lower": True,
        "tl.disable_warp_specialized": True
    })
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
def chunk_linear_attn_bwd_kernel(
    B,
    S,
    H,
    DK,
    DV,
    dtype: str = 'float16',
    scale: float = None,
) -> torch.Tensor:

    if scale is None:
        scale = DK**-0.5
    accum_dtype = 'float'

    chunk_size = 64
31
    BK = BV = 64  # Set to 128 can be faster, but has some numerical differences with FLA
32
33
34
35
36
37
    assert S % chunk_size == 0 and DK % BK == 0 and DV % BV == 0
    NK = tl.cdiv(DK, BK)
    NV = tl.cdiv(DV, BV)
    NT = tl.cdiv(S, chunk_size)

    @T.prim_func
38
39
40
41
42
43
44
45
    def chunk_linear_attn_bwd(
            Q: T.Tensor([B, S, H, DK], dtype),  # type: ignore
            K: T.Tensor([B, S, H, DK], dtype),  # type: ignore
            V: T.Tensor([B, S, H, DV], dtype),  # type: ignore
            dO: T.Tensor([B, S, H, DV], dtype),  # type: ignore
            dQ: T.Tensor([NV, B, S, H, DK], dtype),  # type: ignore
            dK: T.Tensor([NV, B, S, H, DK], dtype),  # type: ignore
            dV: T.Tensor([NK, B, S, H, DV], dtype),  # type: ignore
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
    ):
        with T.Kernel(NV, NK, B * H) as (i_v, i_k, i_bh):
            i_b = i_bh // H
            i_h = i_bh % H

            ds = T.alloc_fragment([chunk_size, chunk_size], accum_dtype)
            ds_shared = T.alloc_shared([chunk_size, chunk_size], dtype)
            dq = T.alloc_fragment([chunk_size, BK], accum_dtype)
            dk = T.alloc_fragment([chunk_size, BK], accum_dtype)
            dv = T.alloc_fragment([chunk_size, BV], accum_dtype)
            q = T.alloc_shared([chunk_size, BK], dtype)
            k = T.alloc_shared([chunk_size, BK], dtype)
            v = T.alloc_shared([chunk_size, BV], dtype)
            do = T.alloc_shared([chunk_size, BV], dtype)
            h = T.alloc_fragment([BV, BK], accum_dtype)
            h_shared = T.alloc_shared([BV, BK], dtype)
            dh = T.alloc_fragment([BK, BV], accum_dtype)
            dh_shared = T.alloc_shared([BK, BV], dtype)
            T.clear(h)
            T.clear(dh)

            T.annotate_layout({
                ds_shared: tl.layout.make_swizzled_layout(ds_shared),
                q: tl.layout.make_swizzled_layout(q),
                k: tl.layout.make_swizzled_layout(k),
                v: tl.layout.make_swizzled_layout(v),
                do: tl.layout.make_swizzled_layout(do),
                h_shared: tl.layout.make_swizzled_layout(h_shared),
                dh_shared: tl.layout.make_swizzled_layout(dh_shared)
            })
76
            T.use_swizzle(10)
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120

            # Calculate dQ
            for i in T.Pipelined(0, NT, num_stages=1):
                T.copy(K[i_b, i * chunk_size:(i + 1) * chunk_size, i_h, i_k * BK:(i_k + 1) * BK], k)
                T.copy(V[i_b, i * chunk_size:(i + 1) * chunk_size, i_h, i_v * BV:(i_v + 1) * BV], v)
                T.copy(dO[i_b, i * chunk_size:(i + 1) * chunk_size, i_h, i_v * BV:(i_v + 1) * BV],
                       do)

                T.gemm(do, v, ds, transpose_B=True, clear_accum=True)
                for row, col in T.Parallel(chunk_size, chunk_size):
                    ds_shared[row, col] = T.if_then_else(row >= col, ds[row, col], 0)

                T.gemm(ds_shared, k, dq, clear_accum=True)
                T.copy(h, h_shared)
                T.gemm(do, h_shared, dq)
                T.gemm(v, k, h, transpose_A=True)
                for row, col in T.Parallel(chunk_size, BK):
                    dq[row, col] *= scale
                T.copy(
                    dq, dQ[i_v, i_b, i * chunk_size:(i + 1) * chunk_size, i_h,
                           i_k * BK:(i_k + 1) * BK])

            # Calculate dK, dV (reversely)
            for i in T.Pipelined(1, NT + 1, num_stages=1):
                start = NT - i
                for row, col in T.Parallel(chunk_size, BK):
                    q[row, col] = Q[i_b, start * chunk_size + row, i_h, i_k * BK + col] * scale
                T.copy(
                    K[i_b, start * chunk_size:(start + 1) * chunk_size, i_h,
                      i_k * BK:(i_k + 1) * BK], k)
                T.copy(
                    V[i_b, start * chunk_size:(start + 1) * chunk_size, i_h,
                      i_v * BV:(i_v + 1) * BV], v)
                T.copy(
                    dO[i_b, start * chunk_size:(start + 1) * chunk_size, i_h,
                       i_v * BV:(i_v + 1) * BV], do)

                # Calculate dk
                T.gemm(
                    v, do, ds, transpose_B=True, clear_accum=True
                )  # ds here actually means `s`, but we simply reuse the buffer `ds`
                for row, col in T.Parallel(chunk_size, chunk_size):
                    ds_shared[row, col] = T.if_then_else(row <= col, ds[row, col], 0)
                T.gemm(ds_shared, q, dk, clear_accum=True)
121
                T.copy(dh, dh_shared)
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
                T.gemm(v, dh_shared, dk, transpose_B=True)

                # Calculate dv
                T.gemm(k, q, ds, transpose_B=True, clear_accum=True)
                for row, col in T.Parallel(chunk_size, chunk_size):
                    ds_shared[row, col] = T.if_then_else(row <= col, ds[row, col], 0)
                T.gemm(ds_shared, do, dv, clear_accum=True)
                T.gemm(k, dh_shared, dv)

                # Update dh
                T.gemm(q, do, dh, transpose_A=True)

                T.copy(
                    dk, dK[i_v, i_b, start * chunk_size:(start + 1) * chunk_size, i_h,
                           i_k * BK:(i_k + 1) * BK])
                T.copy(
                    dv, dV[i_k, i_b, start * chunk_size:(start + 1) * chunk_size, i_h,
                           i_v * BV:(i_v + 1) * BV])

141
    return chunk_linear_attn_bwd
142
143
144
145
146
147
148
149
150
151
152
153


def postprocess(dQ, dK, dV):
    dQ = dQ[0] if dQ.size(0) == 1 else dQ.sum(0)
    dK = dK[0] if dK.size(0) == 1 else dK.sum(0)
    dV = dV[0] if dV.size(0) == 1 else dV.sum(0)
    return dQ, dK, dV


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument('--B', type=int, default=8, help='Batch size')
154
155
    parser.add_argument('--S', type=int, default=4096, help='Seq len')
    parser.add_argument('--H', type=int, default=32, help='Num heads')
156
157
158
159
160
161
162
163
164
    parser.add_argument('--D', type=int, default=256, help='Head dim')
    args = parser.parse_args()
    B, S, H, D = args.B, args.S, args.H, args.D

    q = torch.randn((B, S, H, D), device='cuda', dtype=torch.float16, requires_grad=True)
    k = torch.randn((B, S, H, D), device='cuda', dtype=torch.float16, requires_grad=True)
    v = torch.randn((B, S, H, D), device='cuda', dtype=torch.float16, requires_grad=True)
    do = torch.randn((B, S, H, D), device='cuda', dtype=torch.float16)

165
    kernel = chunk_linear_attn_bwd_kernel(B, S, H, D, D)
166
    dq, dk, dv = postprocess(*kernel(q, k, v, do))
167
    o_ref, _ = fused_chunk_linear_attn(q, k, v, output_final_state=True, normalize=False)
168
169
170
171
172
173
174
    o_ref.backward(do, retain_graph=True)
    if torch.allclose(dq, q.grad) and torch.allclose(dk, k.grad) and torch.allclose(dv, v.grad):
        print('Passed all tests!✅')
    else:
        print('Failed some tests!❌')
    t1 = do_bench(lambda: o_ref.backward(do, retain_graph=True), warmup=25, rep=100)
    q.grad = k.grad = v.grad = None
175
    o_ref, _ = fused_chunk_linear_attn(q, k, v, output_final_state=True, normalize=False)
176
177
178
179
180
181
182
183
    t2 = do_bench(lambda: postprocess(*kernel(q, k, v, do)), warmup=25, rep=100)
    print(f'Triton latency: {t1:.3f} ms')
    print(f'TileLang latency: {t2:.3f} ms')
    print(f'Speedup: {t1/t2:.3f}x')


if __name__ == '__main__':
    main()