Tritonは並列プログラミングのための言語とコンパイラです。カスタムDNN計算カーネルを効率的に記述し、最新のGPUハードウェア上で最大スループットで実行できるようにするためのPythonベースのプログラミング環境を提供するように設計されています。

Triton の中国語ドキュメントの詳細については、→ https://triton.hyper.ai/ をご覧ください。

これはTriDaoのFlash Attention v2アルゴリズムに基づくTriton実装です。謝辞：OpenAI Core Team

特別な感謝を申し上げます:

• フラッシュアテンションに関するオリジナル論文 (https://arxiv.org/abs/2205.14135)
• ラーベとシュターツ (https://arxiv.org/pdf/2112.05682v2.pdf)

外：

融合アテンションバッチ4ヘッド32d64フォワード因果関係=真:

融合注意バッチ4ヘッド32d64転送因果=偽:

融合注意バッチ4ヘッド32D64BWD因果関係=真:

 import pytest import torch import triton import triton.language as tl def is_hip(): return triton.runtime.driver.active.get_current_target().backend == "hip" @triton.jit def _attn_fwd_inner(acc, l_i, m_i, q, # K_block_ptr, V_block_ptr, # start_m, qk_scale, # BLOCK_M: tl.constexpr, HEAD_DIM: tl.constexpr, BLOCK_N: tl.constexpr, # STAGE: tl.constexpr, offs_m: tl.constexpr, offs_n: tl.constexpr, # N_CTX: tl.constexpr, fp8_v: tl.constexpr): # range of values handled by this stage # 这个阶段处理的数值范围if STAGE == 1: lo, hi = 0, start_m * BLOCK_M elif STAGE == 2: lo, hi = start_m * BLOCK_M, (start_m + 1) * BLOCK_M lo = tl.multiple_of(lo, BLOCK_M) # causal = False else: lo, hi = 0, N_CTX K_block_ptr = tl.advance(K_block_ptr, (0, lo)) V_block_ptr = tl.advance(V_block_ptr, (lo, 0)) # loop over k, v and update accumulator # 循环遍历k 和v，并更新累加器。 for start_n in range(lo, hi, BLOCK_N): start_n = tl.multiple_of(start_n, BLOCK_N) # -- compute qk ---- # -- 计算qk ---- k = tl.load(K_block_ptr) qk = tl.dot(q, k) if STAGE == 2: mask = offs_m[:, None] >= (start_n + offs_n[None, :]) qk = qk * qk_scale + tl.where(mask, 0, -1.0e6) m_ij = tl.maximum(m_i, tl.max(qk, 1)) qk -= m_ij[:, None] else: m_ij = tl.maximum(m_i, tl.max(qk, 1) * qk_scale) qk = qk * qk_scale - m_ij[:, None] p = tl.math.exp2(qk) l_ij = tl.sum(p, 1) # -- update m_i and l_i # -- 更新m_i 和l_i alpha = tl.math.exp2(m_i - m_ij) l_i = l_i * alpha + l_ij # -- update output accumulator -- # -- 更新输出累加器-- acc = acc * alpha[:, None] # update acc # 更新acc v = tl.load(V_block_ptr) if fp8_v: p = p.to(tl.float8e5) else: p = p.to(tl.float16) acc = tl.dot(p, v, acc) # update m_i and l_i # 更新m_i 和l_i m_i = m_ij V_block_ptr = tl.advance(V_block_ptr, (BLOCK_N, 0)) K_block_ptr = tl.advance(K_block_ptr, (0, BLOCK_N)) return acc, l_i, m_i # We don't run auto-tuning every time to keep the tutorial fast. Keeping # the code below and commenting out the equivalent parameters is convenient for # re-tuning. # 为了重新调整，我们不会每次都运行自动调优以保持教程的快速性。 # 保留下面的代码并注释掉等效的参数是方便的做法。 configs = [ triton.Config({'BLOCK_M': BM, 'BLOCK_N': BN}, num_stages=s, num_warps=w) \ for BM in [64, 128]\ for BN in [32, 64]\ for s in ([1] if is_hip() else [3, 4, 7])\ for w in [4, 8]\ ] def keep(conf): BLOCK_M = conf.kwargs["BLOCK_M"] BLOCK_N = conf.kwargs["BLOCK_N"] if BLOCK_M * BLOCK_N < 128 * 128 and conf.num_warps == 8: return False return True @triton.autotune(list(filter(keep, configs)), key=["N_CTX", "HEAD_DIM"]) @triton.jit def _attn_fwd(Q, K, V, sm_scale, M, Out, # stride_qz, stride_qh, stride_qm, stride_qk, # stride_kz, stride_kh, stride_kn, stride_kk, # stride_vz, stride_vh, stride_vk, stride_vn, # stride_oz, stride_oh, stride_om, stride_on, # Z, H, N_CTX, # HEAD_DIM: tl.constexpr, # BLOCK_M: tl.constexpr, # BLOCK_N: tl.constexpr, # STAGE: tl.constexpr # ): tl.static_assert(BLOCK_N <= HEAD_DIM) start_m = tl.program_id(0) off_hz = tl.program_id(1) off_z = off_hz // H off_h = off_hz % H qvk_offset = off_z.to(tl.int64) * stride_qz + off_h.to(tl.int64) * stride_qh # block pointers # 块指针Q_block_ptr = tl.make_block_ptr( base=Q + qvk_offset, shape=(N_CTX, HEAD_DIM), strides=(stride_qm, stride_qk), offsets=(start_m * BLOCK_M, 0), block_shape=(BLOCK_M, HEAD_DIM), order=(1, 0), ) v_order: tl.constexpr = (0, 1) if V.dtype.element_ty == tl.float8e5 else (1, 0) V_block_ptr = tl.make_block_ptr( base=V + qvk_offset, shape=(N_CTX, HEAD_DIM), strides=(stride_vk, stride_vn), offsets=(0, 0), block_shape=(BLOCK_N, HEAD_DIM), order=v_order, ) K_block_ptr = tl.make_block_ptr( base=K + qvk_offset, shape=(HEAD_DIM, N_CTX), strides=(stride_kk, stride_kn), offsets=(0, 0), block_shape=(HEAD_DIM, BLOCK_N), order=(0, 1), ) O_block_ptr = tl.make_block_ptr( base=Out + qvk_offset, shape=(N_CTX, HEAD_DIM), strides=(stride_om, stride_on), offsets=(start_m * BLOCK_M, 0), block_shape=(BLOCK_M, HEAD_DIM), order=(1, 0), ) # initialize offsets # 初始化偏移量offs_m = start_m * BLOCK_M + tl.arange(0, BLOCK_M) offs_n = tl.arange(0, BLOCK_N) # initialize pointer to m and l、 # 初始化m 和l 的指针m_i = tl.zeros([BLOCK_M], dtype=tl.float32) - float("inf") l_i = tl.zeros([BLOCK_M], dtype=tl.float32) + 1.0 acc = tl.zeros([BLOCK_M, HEAD_DIM], dtype=tl.float32) # load scales # 加载系数qk_scale = sm_scale qk_scale *= 1.44269504 # 1/log(2) # load q: it will stay in SRAM throughout # 加载q: 它将始终保留在SRAM 中q = tl.load(Q_block_ptr) # stage 1: off-band # 阶段1: off-band # For causal = True, STAGE = 3 and _attn_fwd_inner gets 1 as its STAGE # 对于causal = True，STAGE = 3，而_attn_fwd_inner 的STAGE 参数设定为1。 # For causal = False, STAGE = 1, and _attn_fwd_inner gets 3 as its STAGE # 对于causal = False, STAGE = 1, 而_attn_fwd_inner 的STAGE 参数设定为3。 if STAGE & 1: acc, l_i, m_i = _attn_fwd_inner(acc, l_i, m_i, q, K_block_ptr, V_block_ptr, # start_m, qk_scale, # BLOCK_M, HEAD_DIM, BLOCK_N, # 4 - STAGE, offs_m, offs_n, N_CTX, V.dtype.element_ty == tl.float8e5 # ) # stage 2: on-band # 阶段2: on-band if STAGE & 2: # barrier makes it easier for compielr to schedule the # two loops independently acc, l_i, m_i = _attn_fwd_inner(acc, l_i, m_i, q, K_block_ptr, V_block_ptr, # start_m, qk_scale, # BLOCK_M, HEAD_DIM, BLOCK_N, # 2, offs_m, offs_n, N_CTX, V.dtype.element_ty == tl.float8e5 # ) # epilogue # 结尾m_i += tl.math.log2(l_i) acc = acc / l_i[:, None] m_ptrs = M + off_hz * N_CTX + offs_m tl.store(m_ptrs, m_i) tl.store(O_block_ptr, acc.to(Out.type.element_ty)) @triton.jit def _attn_bwd_preprocess(O, DO, # Delta, # Z, H, N_CTX, # BLOCK_M: tl.constexpr, HEAD_DIM: tl.constexpr # ): off_m = tl.program_id(0) * BLOCK_M + tl.arange(0, BLOCK_M) off_hz = tl.program_id(1) off_n = tl.arange(0, HEAD_DIM) # load # 加载o = tl.load(O + off_hz * HEAD_DIM * N_CTX + off_m[:, None] * HEAD_DIM + off_n[None, :]) do = tl.load(DO + off_hz * HEAD_DIM * N_CTX + off_m[:, None] * HEAD_DIM + off_n[None, :]).to(tl.float32) delta = tl.sum(o * do, axis=1) # write-back # 写回tl.store(Delta + off_hz * N_CTX + off_m, delta) # The main inner-loop logic for computing dK and dV. # 内循环的主要逻辑是计算dK 和dV @triton.jit def _attn_bwd_dkdv(dk, dv, # Q, k, v, sm_scale, # DO, # M, D, # # shared by Q/K/V/DO. # 由Q/K/V/DO 共享stride_tok, stride_d, # H, N_CTX, BLOCK_M1: tl.constexpr, # BLOCK_N1: tl.constexpr, # HEAD_DIM: tl.constexpr, # # Filled in by the wrapper. # 由wrapper 填充start_n, start_m, num_steps, # MASK: tl.constexpr): offs_m = start_m + tl.arange(0, BLOCK_M1) offs_n = start_n + tl.arange(0, BLOCK_N1) offs_k = tl.arange(0, HEAD_DIM) qT_ptrs = Q + offs_m[None, :] * stride_tok + offs_k[:, None] * stride_d do_ptrs = DO + offs_m[:, None] * stride_tok + offs_k[None, :] * stride_d # BLOCK_N1 must be a multiple of BLOCK_M1, otherwise the code wouldn't work. # BLOCK_N1 必须是BLOCK_M1 的倍数，否则代码将无法正常工作。 tl.static_assert(BLOCK_N1 % BLOCK_M1 == 0) curr_m = start_m step_m = BLOCK_M1 for blk_idx in range(num_steps): qT = tl.load(qT_ptrs) # Load m before computing qk to reduce pipeline stall. # 在计算qk 前先加载m，以减少流水线停顿。 offs_m = curr_m + tl.arange(0, BLOCK_M1) m = tl.load(M + offs_m) qkT = tl.dot(k, qT) pT = tl.math.exp2(qkT - m[None, :]) # Autoregressive masking. # 自回归掩码。 if MASK: mask = (offs_m[None, :] >= offs_n[:, None]) pT = tl.where(mask, pT, 0.0) do = tl.load(do_ptrs) # Compute dV. # 计算dV. ppT = pT ppT = ppT.to(tl.float16) dv += tl.dot(ppT, do) # D (= delta) is pre-divided by ds_scale. # D（= delta）已经通过ds_scale 进行了预除。 Di = tl.load(D + offs_m) # Compute dP and dS. # 计算dP 和dS. dpT = tl.dot(v, tl.trans(do)).to(tl.float32) dsT = pT * (dpT - Di[None, :]) dsT = dsT.to(tl.float16) dk += tl.dot(dsT, tl.trans(qT)) # Increment pointers. # 增加指针。 curr_m += step_m qT_ptrs += step_m * stride_tok do_ptrs += step_m * stride_tok return dk, dv # the main inner-loop logic for computing dQ # 内循环的主要逻辑是计算dQ @triton.jit def _attn_bwd_dq(dq, q, K, V, # do, m, D, # shared by Q/K/V/DO. # 由Q/K/V/DO 共享stride_tok, stride_d, # H, N_CTX, # BLOCK_M2: tl.constexpr, # BLOCK_N2: tl.constexpr, # HEAD_DIM: tl.constexpr, # Filled in by the wrapper. # 由wrapper 填充start_m, start_n, num_steps, # MASK: tl.constexpr): offs_m = start_m + tl.arange(0, BLOCK_M2) offs_n = start_n + tl.arange(0, BLOCK_N2) offs_k = tl.arange(0, HEAD_DIM) kT_ptrs = K + offs_n[None, :] * stride_tok + offs_k[:, None] * stride_d vT_ptrs = V + offs_n[None, :] * stride_tok + offs_k[:, None] * stride_d # D (= delta) is pre-divided by ds_scale. # D（即delta）在使用前已经被ds_scale 预先除以。 Di = tl.load(D + offs_m) # BLOCK_M2 must be a multiple of BLOCK_N2, otherwise the code wouldn't work. # BLOCK_M2 必须是BLOCK_N2 的倍数，否则代码将无法正常工作。 tl.static_assert(BLOCK_M2 % BLOCK_N2 == 0) curr_n = start_n step_n = BLOCK_N2 for blk_idx in range(num_steps): kT = tl.load(kT_ptrs) vT = tl.load(vT_ptrs) qk = tl.dot(q, kT) p = tl.math.exp2(qk - m) # Autoregressive masking. # 自回归掩码。 if MASK: offs_n = curr_n + tl.arange(0, BLOCK_N2) mask = (offs_m[:, None] >= offs_n[None, :]) p = tl.where(mask, p, 0.0) # Compute dP and dS. # 计算dP 和dS. dp = tl.dot(do, vT).to(tl.float32) ds = p * (dp - Di[:, None]) ds = ds.to(tl.float16) # Compute dQ. # 计算dQ. # NOTE: We need to de-scale dq in the end, because kT was pre-scaled. # 注意：我们需要在最后对dq 进行反缩放，因为kT 已经进行了预缩放。 dq += tl.dot(ds, tl.trans(kT)) # 增加指针。 curr_n += step_n kT_ptrs += step_n * stride_tok vT_ptrs += step_n * stride_tok return dq @triton.jit def _attn_bwd(Q, K, V, sm_scale, # DO, # DQ, DK, DV, # M, D, # shared by Q/K/V/DO. # 由Q/K/V/DO 共享stride_z, stride_h, stride_tok, stride_d, # H, N_CTX, # BLOCK_M1: tl.constexpr, # BLOCK_N1: tl.constexpr, # BLOCK_M2: tl.constexpr, # BLOCK_N2: tl.constexpr, # BLK_SLICE_FACTOR: tl.constexpr, # HEAD_DIM: tl.constexpr): LN2: tl.constexpr = 0.6931471824645996 # = ln(2) bhid = tl.program_id(2) off_chz = (bhid * N_CTX).to(tl.int64) adj = (stride_h * (bhid % H) + stride_z * (bhid // H)).to(tl.int64) pid = tl.program_id(0) # offset pointers for batch/head # 为batch/head 偏移指针。 Q += adj K += adj V += adj DO += adj DQ += adj DK += adj DV += adj M += off_chz D += off_chz # load scales # 加载系数offs_k = tl.arange(0, HEAD_DIM) start_n = pid * BLOCK_N1 start_m = start_n MASK_BLOCK_M1: tl.constexpr = BLOCK_M1 // BLK_SLICE_FACTOR offs_n = start_n + tl.arange(0, BLOCK_N1) dv = tl.zeros([BLOCK_N1, HEAD_DIM], dtype=tl.float32) dk = tl.zeros([BLOCK_N1, HEAD_DIM], dtype=tl.float32) # load K and V: they stay in SRAM throughout the inner loop. # 加载K 和V：它们在整个内部循环过程中保持在SRAM 中。 k = tl.load(K + offs_n[:, None] * stride_tok + offs_k[None, :] * stride_d) v = tl.load(V + offs_n[:, None] * stride_tok + offs_k[None, :] * stride_d) num_steps = BLOCK_N1 // MASK_BLOCK_M1 dk, dv = _attn_bwd_dkdv(dk, dv, # Q, k, v, sm_scale, # DO, # M, D, # stride_tok, stride_d, # H, N_CTX, # MASK_BLOCK_M1, BLOCK_N1, HEAD_DIM, # start_n, start_m, num_steps, # MASK=True # ) start_m += num_steps * MASK_BLOCK_M1 num_steps = (N_CTX - start_m) // BLOCK_M1 # Compute dK and dV for non-masked blocks. # 计算非掩码块的dK 和dV。 dk, dv = _attn_bwd_dkdv( # dk, dv, # Q, k, v, sm_scale, # DO, # M, D, # stride_tok, stride_d, # H, N_CTX, # BLOCK_M1, BLOCK_N1, HEAD_DIM, # start_n, start_m, num_steps, # MASK=False # ) dv_ptrs = DV + offs_n[:, None] * stride_tok + offs_k[None, :] * stride_d tl.store(dv_ptrs, dv) # Write back dK. # 写回dK dk *= sm_scale dk_ptrs = DK + offs_n[:, None] * stride_tok + offs_k[None, :] * stride_d tl.store(dk_ptrs, dk) # THIS BLOCK DOES DQ: # 该段计算DQ start_m = pid * BLOCK_M2 end_n = start_m + BLOCK_M2 MASK_BLOCK_N2: tl.constexpr = BLOCK_N2 // BLK_SLICE_FACTOR offs_m = start_m + tl.arange(0, BLOCK_M2) q = tl.load(Q + offs_m[:, None] * stride_tok + offs_k[None, :] * stride_d) dq = tl.zeros([BLOCK_M2, HEAD_DIM], dtype=tl.float32) do = tl.load(DO + offs_m[:, None] * stride_tok + offs_k[None, :] * stride_d) m = tl.load(M + offs_m) m = m[:, None] # Compute dQ for masked (diagonal) blocks. # 计算掩码（对角线）块的dQ。 # NOTE: This code scans each row of QK^T backward (from right to left, # but inside each call to _attn_bwd_dq, from left to right), but that's # not due to anything important. I just wanted to reuse the loop # structure for dK & dV above as much as possible. # 注意：这段代码逆向扫描每行QK^T（从右向左，但在每次调用_attn_bwd_dq 内部， # 从左向右），但这并不重要。我只是希望尽可能多地重用上述dK 和dV 的循环结构。 num_steps = BLOCK_M2 // MASK_BLOCK_N2 dq = _attn_bwd_dq(dq, q, K, V, # do, m, D, # stride_tok, stride_d, # H, N_CTX, # BLOCK_M2, MASK_BLOCK_N2, HEAD_DIM, # start_m, end_n - num_steps * MASK_BLOCK_N2, num_steps, # MASK=True # ) end_n -= num_steps * MASK_BLOCK_N2 # stage 2 # 阶段2 num_steps = end_n // BLOCK_N2 dq = _attn_bwd_dq(dq, q, K, V, # do, m, D, # stride_tok, stride_d, # H, N_CTX, # BLOCK_M2, BLOCK_N2, HEAD_DIM, # start_m, end_n - num_steps * BLOCK_N2, num_steps, # MASK=False # ) # Write back dQ. # 写回dQ dq_ptrs = DQ + offs_m[:, None] * stride_tok + offs_k[None, :] * stride_d dq *= LN2 tl.store(dq_ptrs, dq) class _attention(torch.autograd.Function): @staticmethod def forward(ctx, q, k, v, causal, sm_scale): # shape constraints # 形状约束HEAD_DIM_Q, HEAD_DIM_K = q.shape[-1], k.shape[-1] # when v is in float8_e5m2 it is transposed. # 当v 在float8_e5m2 格式下时，进行转置。 HEAD_DIM_V = v.shape[-1] assert HEAD_DIM_Q == HEAD_DIM_K and HEAD_DIM_K == HEAD_DIM_V assert HEAD_DIM_K in {16, 32, 64, 128, 256} o = torch.empty_like(q) stage = 3 if causal else 1 extra_kern_args = {} # Tuning for AMD target # 为AMD 设备调整if is_hip(): waves_per_eu = 3 if HEAD_DIM_K <= 64 else 2 extra_kern_args = {"waves_per_eu": waves_per_eu, "allow_flush_denorm": True} grid = lambda args: (triton.cdiv(q.shape[2], args["BLOCK_M"]), q.shape[0] * q.shape[1], 1) M = torch.empty((q.shape[0], q.shape[1], q.shape[2]), device=q.device, dtype=torch.float32) _attn_fwd[grid]( q, k, v, sm_scale, M, o, # q.stride(0), q.stride(1), q.stride(2), q.stride(3), # k.stride(0), k.stride(1), k.stride(2), k.stride(3), # v.stride(0), v.stride(1), v.stride(2), v.stride(3), # o.stride(0), o.stride(1), o.stride(2), o.stride(3), # q.shape[0], q.shape[1], # N_CTX=q.shape[2], # HEAD_DIM=HEAD_DIM_K, # STAGE=stage, # **extra_kern_args) ctx.save_for_backward(q, k, v, o, M) ctx.grid = grid ctx.sm_scale = sm_scale ctx.HEAD_DIM = HEAD_DIM_K ctx.causal = causal return o @staticmethod def backward(ctx, do): q, k, v, o, M = ctx.saved_tensors assert do.is_contiguous() assert q.stride() == k.stride() == v.stride() == o.stride() == do.stride() dq = torch.empty_like(q) dk = torch.empty_like(k) dv = torch.empty_like(v) BATCH, N_HEAD, N_CTX = q.shape[:3] PRE_BLOCK = 128 NUM_WARPS, NUM_STAGES = 4, 5 BLOCK_M1, BLOCK_N1, BLOCK_M2, BLOCK_N2 = 32, 128, 128, 32 BLK_SLICE_FACTOR = 2 RCP_LN2 = 1.4426950408889634 # = 1.0 / ln(2) arg_k = k arg_k = arg_k * (ctx.sm_scale * RCP_LN2) PRE_BLOCK = 128 assert N_CTX % PRE_BLOCK == 0 pre_grid = (N_CTX // PRE_BLOCK, BATCH * N_HEAD) delta = torch.empty_like(M) _attn_bwd_preprocess[pre_grid]( o, do, # delta, # BATCH, N_HEAD, N_CTX, # BLOCK_M=PRE_BLOCK, HEAD_DIM=ctx.HEAD_DIM # ) grid = (N_CTX // BLOCK_N1, 1, BATCH * N_HEAD) _attn_bwd[grid]( q, arg_k, v, ctx.sm_scale, do, dq, dk, dv, # M, delta, # q.stride(0), q.stride(1), q.stride(2), q.stride(3), # N_HEAD, N_CTX, # BLOCK_M1=BLOCK_M1, BLOCK_N1=BLOCK_N1, # BLOCK_M2=BLOCK_M2, BLOCK_N2=BLOCK_N2, # BLK_SLICE_FACTOR=BLK_SLICE_FACTOR, # HEAD_DIM=ctx.HEAD_DIM, # num_warps=NUM_WARPS, # num_stages=NUM_STAGES # ) return dq, dk, dv, None, None attention = _attention.apply @pytest.mark.parametrize("Z, H, N_CTX, HEAD_DIM", [(1, 2, 1024, 64)]) @pytest.mark.parametrize("causal", [True]) def test_op(Z, H, N_CTX, HEAD_DIM, causal, dtype=torch.float16): torch.manual_seed(20) q = (torch.empty((Z, H, N_CTX, HEAD_DIM), dtype=dtype, device="cuda").normal_(mean=0.0, std=0.5).requires_grad_()) k = (torch.empty((Z, H, N_CTX, HEAD_DIM), dtype=dtype, device="cuda").normal_(mean=0.0, std=0.5).requires_grad_()) v = (torch.empty((Z, H, N_CTX, HEAD_DIM), dtype=dtype, device="cuda").normal_(mean=0.0, std=0.5).requires_grad_()) sm_scale = 0.5 dout = torch.randn_like(q) # reference implementation # 参考实现M = torch.tril(torch.ones((N_CTX, N_CTX), device="cuda")) p = torch.matmul(q, k.transpose(2, 3)) * sm_scale if causal: p[:, :, M == 0] = float("-inf") p = torch.softmax(p.float(), dim=-1).half() # p = torch.exp(p) ref_out = torch.matmul(p, v) ref_out.backward(dout) ref_dv, v.grad = v.grad.clone(), None ref_dk, k.grad = k.grad.clone(), None ref_dq, q.grad = q.grad.clone(), None # triton implementation # triton 实现tri_out = attention(q, k, v, causal, sm_scale).half() tri_out.backward(dout) tri_dv, v.grad = v.grad.clone(), None tri_dk, k.grad = k.grad.clone(), None tri_dq, q.grad = q.grad.clone(), None # compare # 比较assert torch.allclose(ref_out, tri_out, atol=1e-2, rtol=0) rtol = 0.0 # Relative tolerance workaround for known hardware limitation of MI200 GPU. # 针对MI200 GPU 已知的硬件限制，使用相对容差的解决方法。 # For details see https://pytorch.org/docs/stable/notes/numerical_accuracy.html#reduced-precision-fp16-and-bf16-gemms-and-convolutions-on-amd-instinct-mi200-devices # 详情见https://pytorch.org/docs/stable/notes/numerical_accuracy.html#reduced-precision-fp16-and-bf16-gemms-and-convolutions-on-amd-instinct-mi200-devices if torch.version.hip is not None and triton.runtime.driver.active.get_current_target().arch == "gfx90a": rtol = 1e-2 assert torch.allclose(ref_dv, tri_dv, atol=1e-2, rtol=rtol) assert torch.allclose(ref_dk, tri_dk, atol=1e-2, rtol=rtol) assert torch.allclose(ref_dq, tri_dq, atol=1e-2, rtol=rtol) try: from flash_attn.flash_attn_interface import \ flash_attn_qkvpacked_func as flash_attn_func HAS_FLASH = True except BaseException: HAS_FLASH = False TORCH_HAS_FP8 = hasattr(torch, 'float8_e5m2') BATCH, N_HEADS, HEAD_DIM = 4, 32, 64 # vary seq length for fixed head and batch=4 # 对于固定的head 和batch 为4，变化序列长度。 configs = [] for mode in ["fwd", "bwd"]: for causal in [True, False]: if mode == "bwd" and not causal: continue configs.append( triton.testing.Benchmark( x_names=["N_CTX"], x_vals=[2**i for i in range(10, 15)], line_arg="provider", line_vals=["triton-fp16"] + (["triton-fp8"] if TORCH_HAS_FP8 else []) + (["flash"] if HAS_FLASH else []), line_names=["Triton [FP16]"] + (["Triton [FP8]"] if TORCH_HAS_FP8 else []) + (["Flash-2"] if HAS_FLASH else []), styles=[("red", "-"), ("blue", "-")], ylabel="ms", plot_name=f"fused-attention-batch{BATCH}-head{N_HEADS}-d{HEAD_DIM}-{mode}-causal={causal}", args={ "H": N_HEADS, "BATCH": BATCH, "HEAD_DIM": HEAD_DIM, "mode": mode, "causal": causal, }, )) @triton.testing.perf_report(configs) def bench_flash_attention(BATCH, H, N_CTX, HEAD_DIM, causal, mode, provider, device="cuda"): assert mode in ["fwd", "bwd"] warmup = 25 rep = 100 dtype = torch.float16 if "triton" in provider: q = torch.randn((BATCH, H, N_CTX, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) k = torch.randn((BATCH, H, N_CTX, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) v = torch.randn((BATCH, H, N_CTX, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) if mode == "fwd" and "fp8" in provider: q = q.to(torch.sfloat8_e5m2) k = k.to(torch.float8_e5m2) v = v.permute(0, 1, 3, 2).contiguous() v = v.permute(0, 1, 3, 2) v = v.to(torch.float8_e5m2) sm_scale = 1.3 fn = lambda: attention(q, k, v, causal, sm_scale) if mode == "bwd": o = fn() do = torch.randn_like(o) fn = lambda: o.backward(do, retain_graph=True) ms = triton.testing.do_bench(fn, warmup=warmup, rep=rep) if provider == "flash": qkv = torch.randn((BATCH, N_CTX, 3, H, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) fn = lambda: flash_attn_func(qkv, causal=causal) if mode == "bwd": o = fn() do = torch.randn_like(o) fn = lambda: o.backward(do, retain_graph=True) ms = triton.testing.do_bench(fn, warmup=warmup, rep=rep) flops_per_matmul = 2.0 * BATCH * H * N_CTX * N_CTX * HEAD_DIM total_flops = 2 * flops_per_matmul if causal: total_flops *= 0.5 if mode == "bwd": total_flops *= 2.5 # 2.0(bwd) + 0.5(recompute) return total_flops / ms * 1e-9 if __name__ == "__main__": # only works on post-Ampere GPUs right now # 目前只适用于安培架构GPU。 bench_flash_attention.run(save_path=".", print_data=True)

Jupyterノートブックをダウンロード: 06-fused-attention.ipynb

Pythonソースコードをダウンロード: 06-fused-attention.py

圧縮ファイルをダウンロード: 06-fused-attention.zip