[KDA] Fuse dAqk and dv

fla/ops/kda/chunk.py

@@ -4,9 +4,8 @@
 
 from fla.modules.l2norm import l2norm_bwd, l2norm_fwd
 from fla.ops.common.chunk_delta_h import chunk_gated_delta_rule_bwd_dhu, chunk_gated_delta_rule_fwd_h
-from fla.ops.common.chunk_o import chunk_bwd_dv_local
-from fla.ops.gla.chunk import chunk_gla_bwd_dA, chunk_gla_fwd_o_gk
-from fla.ops.kda.chunk_inter import chunk_kda_bwd_dqkwg
+from fla.ops.gla.chunk import chunk_gla_fwd_o_gk
+from fla.ops.kda.chunk_bwd import chunk_kda_bwd_dAv, chunk_kda_bwd_dqkwg
 from fla.ops.kda.chunk_intra import chunk_kda_bwd_intra, chunk_kda_fwd_intra
 from fla.ops.kda.gate import kda_gate_bwd, kda_gate_fwd
 from fla.ops.kda.wy_fast import prepare_wy_repr_bwd, recompute_w_u_fwd
@@ -103,9 +102,12 @@ def chunk_kda_bwd(
         chunk_indices=chunk_indices,
         use_exp2=True,
     )
-    dv = chunk_bwd_dv_local(
+    # dAqk = do @ v.T
+    # dv = A @ do
+    dAqk, dv = chunk_kda_bwd_dAv(
         q=q,
         k=k,
+        v=v_new,
         do=do,
         A=Aqk,
         scale=scale,
@@ -128,15 +130,6 @@ def chunk_kda_bwd(
         chunk_indices=chunk_indices,
         use_exp2=True,
     )
-    # dq dk in fp32
-    dAqk = chunk_gla_bwd_dA(
-        v=v_new,
-        do=do,
-        scale=scale,
-        cu_seqlens=cu_seqlens,
-        chunk_size=chunk_size,
-        chunk_indices=chunk_indices,
-    )
     dq, dk, dw, dg = chunk_kda_bwd_dqkwg(
         q=q,
         k=k,

fla/ops/kda/chunk_inter.py → fla/ops/kda/chunk_bwd.py

@@ -1,16 +1,94 @@
 # Copyright (c) 2023-2025, Songlin Yang, Yu Zhang
 
-
 import torch
 import triton
 import triton.language as tl
 
 from fla.ops.utils import prepare_chunk_indices
 from fla.ops.utils.op import exp2
-from fla.utils import autotune_cache_kwargs, check_shared_mem
+from fla.utils import IS_NVIDIA_HOPPER, autotune_cache_kwargs, check_shared_mem
 
 BK_LIST = [32, 64] if check_shared_mem() else [16, 32]
 BV_LIST = [64, 128] if check_shared_mem('ampere') else [16, 32]
+NUM_WARPS = [2, 4] if IS_NVIDIA_HOPPER else [2, 4, 8]
+
+
+@triton.heuristics({
+    'IS_VARLEN': lambda args: args['cu_seqlens'] is not None,
+})
+@triton.autotune(
+    configs=[
+        triton.Config({}, num_warps=num_warps, num_stages=num_stages)
+        for num_warps in NUM_WARPS
+        for num_stages in [2, 3, 4]
+    ],
+    key=['H', 'K', 'V', 'BT', 'BK', 'BV'],
+    **autotune_cache_kwargs,
+)
+@triton.jit(do_not_specialize=['T'])
+def chunk_bwd_kernel_dAv(
+    q,
+    k,
+    v,
+    A,
+    do,
+    dv,
+    dA,
+    cu_seqlens,
+    chunk_indices,
+    scale,
+    T,
+    H: tl.constexpr,
+    K: tl.constexpr,
+    V: tl.constexpr,
+    BT: tl.constexpr,
+    BK: tl.constexpr,
+    BV: tl.constexpr,
+    IS_VARLEN: tl.constexpr,
+):
+    i_t, i_bh = tl.program_id(0), tl.program_id(1)
+    i_b, i_h = i_bh // H, i_bh % H
+    if IS_VARLEN:
+        i_n, i_t = tl.load(chunk_indices + i_t * 2).to(tl.int32), tl.load(chunk_indices + i_t * 2 + 1).to(tl.int32)
+        bos, eos = tl.load(cu_seqlens + i_n).to(tl.int32), tl.load(cu_seqlens + i_n + 1).to(tl.int32)
+        T = eos - bos
+    else:
+        bos, eos = i_b * T, i_b * T + T
+
+    # offset calculation
+    q += (bos * H + i_h) * K
+    k += (bos * H + i_h) * K
+    v += (bos * H + i_h) * V
+    do += (bos * H + i_h) * V
+    dv += (bos * H + i_h) * V
+    dA += (bos * H + i_h) * BT
+
+    p_A = tl.make_block_ptr(A + (bos * H + i_h) * BT, (BT, T), (1, H*BT), (0, i_t * BT), (BT, BT), (0, 1))
+    b_A = tl.load(p_A, boundary_check=(0, 1))
+
+    o_t = i_t * BT + tl.arange(0, BT)
+    m_t = o_t < T
+    m_A = (o_t[:, None] <= o_t[None, :]) & (m_t[:, None] & m_t)
+    b_A = tl.where(m_A, b_A, 0).to(do.dtype.element_ty)
+
+    b_dA = tl.zeros([BT, BT], dtype=tl.float32)
+    for i_v in range(tl.cdiv(V, BV)):
+        p_v = tl.make_block_ptr(v, (V, T), (1, H*V), (i_v * BV, i_t * BT), (BV, BT), (0, 1))
+        p_do = tl.make_block_ptr(do, (T, V), (H*V, 1), (i_t * BT, i_v * BV), (BT, BV), (1, 0))
+        p_dv = tl.make_block_ptr(dv, (T, V), (H*V, 1), (i_t * BT, i_v * BV), (BT, BV), (1, 0))
+        # [BV, BT]
+        b_v = tl.load(p_v, boundary_check=(0, 1))
+        # [BT, BV]
+        b_do = tl.load(p_do, boundary_check=(0, 1))
+        # [BT, BT]
+        b_dA += tl.dot(b_do, b_v)
+        # [BT, BV]
+        b_dv = tl.dot(b_A.to(b_do.dtype), b_do)
+        tl.store(p_dv, b_dv.to(p_dv.dtype.element_ty), boundary_check=(0, 1))
+
+    p_dA = tl.make_block_ptr(dA, (T, BT), (H*BT, 1), (i_t * BT, 0), (BT, BT), (1, 0))
+    b_dA = tl.where(o_t[:, None] >= o_t, b_dA * scale, 0.)
+    tl.store(p_dA, b_dA.to(p_dA.dtype.element_ty), boundary_check=(0, 1))
 
 
 @triton.heuristics({
@@ -28,7 +106,7 @@
     **autotune_cache_kwargs,
 )
 @triton.jit(do_not_specialize=['T'])
-def chunk_kda_bwd_kernel_inter(
+def chunk_kda_bwd_kernel_dqkwg(
     q,
     k,
     v,
@@ -138,6 +216,57 @@ def chunk_kda_bwd_kernel_inter(
     tl.store(p_dg, b_dg.to(p_dg.dtype.element_ty), boundary_check=(0, 1))
 
 
+def chunk_kda_bwd_dAv(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    do: torch.Tensor,
+    A: torch.Tensor | None = None,
+    scale: float = None,
+    cu_seqlens: torch.LongTensor | None = None,
+    chunk_size: int = 64,
+    chunk_indices: torch.LongTensor | None = None,
+) -> torch.Tensor:
+    B, T, H, K, V = *k.shape, do.shape[-1]
+    BT = chunk_size
+    if chunk_indices is None and cu_seqlens is not None:
+        chunk_indices = prepare_chunk_indices(cu_seqlens, BT)
+    # H100 can have larger block size
+    if check_shared_mem('hopper', k.device.index):
+        CONST_TILING = 128
+    elif check_shared_mem:
+        CONST_TILING = 64
+    else:
+        CONST_TILING = 32
+    BK = min(max(triton.next_power_of_2(K), 16), CONST_TILING)
+    BV = min(max(triton.next_power_of_2(V), 16), CONST_TILING)
+    NT = triton.cdiv(T, BT) if cu_seqlens is None else len(chunk_indices)
+
+    dA = v.new_empty(B, T, H, BT, dtype=torch.float)
+    dv = torch.empty_like(do)
+    grid = (NT, B * H)
+    chunk_bwd_kernel_dAv[grid](
+        q=q,
+        k=k,
+        v=v,
+        A=A,
+        do=do,
+        dv=dv,
+        dA=dA,
+        cu_seqlens=cu_seqlens,
+        chunk_indices=chunk_indices,
+        scale=scale,
+        T=T,
+        H=H,
+        K=K,
+        V=V,
+        BT=BT,
+        BK=BK,
+        BV=BV,
+    )
+    return dA, dv
+
+
 def chunk_kda_bwd_dqkwg(
     q: torch.Tensor,
     k: torch.Tensor,
@@ -165,7 +294,7 @@ def chunk_kda_bwd_dqkwg(
     dw = torch.empty_like(w)
     dg = torch.empty_like(g)
     def grid(meta): return (triton.cdiv(K, meta['BK']), NT, B * H)
-    chunk_kda_bwd_kernel_inter[grid](
+    chunk_kda_bwd_kernel_dqkwg[grid](
         q=q,
         k=k,
         v=v,