3Kiu'0SrSSKrSSKJr SSKJrJr SSKr\R"\ 5r SS/r \"SS9S \ S \ 4S j5r"S S\5r\R R#S 0S9S+S\R$S\R$S\R$S\R$S\R$S\ S\ S\ S \\R$\R$\R$44Sjj5r\R*S+S\R$S\R$S\R$S\R$S\R$S\ S\ S\ S \\R$\R$\R$44Sjj5rS,S\R$S\R$S\R$S\R$S\R$S\ S\ S\ S\S-S \R$\\R$\R$4-4SjjrS\S\\S4S\S S4Sjr\R R#S 0S9S!\R$S\R$S\R$S\R$S"\R$S#\R$S\R$S\R$S\ S\ S\ S$\R$S \\R$\R$\R$44S%j5r\R*S!\R$S\R$S\R$S\R$S"\R$S#\R$S\R$S\R$S\ S\ S\ S$\R$S \\R$\R$\R$44S&j5rS\S!\R$S'\R$S(\R$S \\R$S-S44 S)jr\R9\\S*9 g)-z Variable-length attention implementation using Flash Attention. This module provides a high-level Python interface for variable-length attention that calls into the optimized Flash Attention kernels. N) lru_cache)Any NamedTuple varlen_attn AuxRequest)maxsize device_indexreturncg)z;Cache device capability check to avoid repeated CUDA calls.F)r s [/var/www/html/dynamic-report/venv/lib/python3.13/site-packages/torch/nn/attention/varlen.py_should_use_cudnnrs c(\rSrSr%SrSr\\S'Srg)rz Request which auxiliary outputs to compute from varlen_attn. Each field is a boolean indicating whether that auxiliary output should be computed. Flser N) __name__ __module__ __qualname____firstlineno____doc__rbool__annotations____static_attributes__r rrrrs Crztorch_attn::_varlen_attn) mutates_argsquerykeyvaluecu_seq_qcu_seq_kmax_qmax_k is_causalcUR=(a [URR5nU(aX[R S5 [ RRRUUUSUUUUSSUS5 n U SU SU Spn OJ[R S 5 [ RRRUUUUUUUSUSS 9 upn n [ R"S [ RURS 9nXU4$) z Private custom op for variable-length attention. This is the internal implementation. Users should use the public varlen_attn function instead. #Using cuDNN backend for varlen_attnNTFr-Using Flash Attention backend for varlen_attn)return_debug_maskdtypedevice) is_cudarr0indexloginfotorchopsaten_cudnn_attention_forward_flash_attention_forwardzerosuint64)rrrr r!r"r#r$ use_cudnnresultoutput softmax_lse rng_state_ rng_state_s r _varlen_attnrC$s" G"3ELL4F4F"GI 6788              *0F1IvayYY @A/4yy~~/V/V         #0W 0 ,Y1 ELLJ  **rc4[R"U5nURS5n URS5n [R"X4[RUR S9n [R"S[R UR S9n XU 4$)z Fake implementation for meta tensor computation and tracing. Based on the 3D varlen path from meta__flash_attention_forward: - query shape: (total, num_heads, head_dim) - logsumexp shape: (num_heads, total_q) rr(r.r,)r5 empty_likesizeemptyfloatr0r;) rrrr r!r"r#r$r>total_q num_heads logsumexpr@s r_varlen_attn_fakerL^sw&  e $FjjmG 1 I EKK I D U\\JI i ''r return_auxc [RRRXX#XEXg5upn UbUR(aX4$U $)a] Compute variable-length attention using Flash Attention. This function is similar to scaled_dot_product_attention but optimized for variable-length sequences using cumulative sequence position tensors. Args: - query (Tensor): Query tensor; shape :math:`(T_q, H, D)` - key (Tensor): Key tensor; shape :math:`(T_k, H, D)` - value (Tensor): Value tensor; shape :math:`(T_k, H, D)` - cu_seq_q (Tensor): Cumulative sequence positions for queries; shape :math:`(N+1,)` - cu_seq_k (Tensor): Cumulative sequence positions for keys/values; shape :math:`(N+1,)` - max_q (int): Maximum query sequence length in the batch. - max_k (int): Maximum key/value sequence length in the batch. - is_causal (bool, optional): If set to True, applies causal masking (default: False). - return_aux (Optional[AuxRequest]): If not None and ``return_aux.lse`` is True, also returns the logsumexp tensor. Shape legend: - :math:`N`: Batch size - :math:`T_q`: Total number of query tokens in the batch (sum of all query sequence lengths) - :math:`T_k`: Total number of key/value tokens in the batch (sum of all key/value sequence lengths) - :math:`H`: Number of attention heads - :math:`D`: Head dimension Returns: - Tensor: Output tensor from attention computation - If ``return_aux`` is not None and ``return_aux.lse`` is True, returns a tuple of Tensors: (output, lse), where lse is the logsumexp Example:: >>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA) >>> batch_size, max_seq_len, embed_dim, num_heads = 2, 512, 1024, 16 >>> head_dim = embed_dim // num_heads >>> seq_lengths = [] >>> for _ in range(batch_size): ... length = torch.randint(1, max_seq_len // 64 + 1, (1,)).item() * 64 ... seq_lengths.append(min(length, max_seq_len)) >>> seq_lengths = torch.tensor(seq_lengths, device="cuda") >>> total_tokens = seq_lengths.sum().item() >>> >>> # Create packed query, key, value tensors >>> query = torch.randn( ... total_tokens, num_heads, head_dim, dtype=torch.float16, device="cuda" ... ) >>> key = torch.randn( ... total_tokens, num_heads, head_dim, dtype=torch.float16, device="cuda" ... ) >>> value = torch.randn( ... total_tokens, num_heads, head_dim, dtype=torch.float16, device="cuda" ... ) >>> >>> # Build cumulative sequence tensor >>> cu_seq = torch.zeros(batch_size + 1, device="cuda", dtype=torch.int32) >>> cu_seq[1:] = seq_lengths.cumsum(0) >>> max_len = seq_lengths.max().item() >>> >>> # Call varlen_attn >>> output = varlen_attn( ... query, key, value, cu_seq, cu_seq, max_len, max_len, is_causal=False ... ) )r5r6 torch_attnrCr) rrrr r!r"r#r$rMoutrrAs rrrsFN))&&33 EXeKCa*..x Jrctxinputs.r>c hUup4pVpxpUupn URX4XVX{X5 XlXlXlgN)save_for_backwardr"r#r$)rQrRr>rrrr r!r"r#r$rPrr@s r_setup_contextrVs=EKBE% Ci%excUIIMrz!torch_attn::_varlen_attn_backwardgrad_outrPrr@c [R"SURS9n UR=(a [ URR 5n U (aO[ RS5 [RRRUUUUUUUUUU SU U U 5upnON[ RS5 [RRRUUUUUUUUUU SU U U 5upnXU4$)Nr)r0r&r'r*) r5rGr0r1rr2r3r4r6r7_cudnn_attention_backward_flash_attention_backward)rWrrrrPrr r!r"r#r$r@unusedr<dqdkdvs r_varlen_attn_backwardr_s[[5<< 0F G"3ELL4F4F"GI 67YY^^==               " @AYY^^==                2:rc [R"U5n [R"U5n [R"U5nXU4$)z> Fake implementation for meta tensor computation and tracing. )r5rE)rWrrrrPrr r!r"r#r$r@ grad_querygrad_key grad_values r_varlen_attn_backward_fakerds?&!!%(J$H!!%(J  ++rgrad_lsegrad_rngcURupEpgppURn URn URn[R R RUUUUU U UUU U UU 5 unnnUUUSSSSSS4 $rT) saved_tensorsr"r#r$r5r6rOr_)rQrWrerfrrrr r!rPrr@r"r#r$r\r]r^s r _backwardri,sBEARAR>E IIE IIE I%%;;        JBB r2tT4tT 99r) setup_context)F)FN)rlogging functoolsrtypingrrr5 getLoggerrr3__all__intrrrlibrary custom_opTensortuplerC register_fakerLrrVr_rdriregister_autogradr rrrws" ! , ' 1CD 3"E6+ <<6+ 6+ <<6+ll 6+ ll 6+  6+ 6+6+ 5<<u|| 346+F6+r( <<( ( <<(ll ( ll (  ( (( 5<<u|| 34((P$(L <<L L <<Lll L ll L  L LLT!L \\E%,, 455L^U38_cd<2N5ll5 <<5 5 << 5  5  5ll5ll5 5 55||5 5<<u|| 345O5p$$,ll, <<, , << ,  ,  ,ll,ll, , ,,||, 5<<u|| 34,%,2: : :05 :HM : 5<<$  #$:4 yGr