o i @sjddlZddlmZddlmZ      ddejjdejdejdejd ejdBd ed ejfd d ZdS)N)PagedAttentionCache)!lazy_import_paged_flash_attentionmoduleqkvattention_maskcachereturnc Kst|jj} t|ddsdn|jddf} | dkrdnd} |dur2|j|||j| d | d d \}}t|tr?|| }| | } d | vrJd | d ini}| | dd  d | | | tj| tj|| f|jd| d|}t|tr|d}|dfS)aPerform the forward pass of attention with paged key-value cache. This function handles the cache updates and performs the attention computation using the flash_attn_varlen_func for efficient processing. Args: q: (total_q, nheads, headdim), where total_q = total number of query tokens in the batch. k: (total_k, nheads_k, headdim), where total_k = total number of key tokens in the batch. but if there is a block table it can be the full k v: (total_k, nheads_k, headdim), where total_k = total number of key tokens in the batch. but if there is a block table it can be the full v cu_seq_lens_q: (batch_size + 1,), dtype torch.int32. The cumulative sequence lengths of the sequences in the batch, used to index into q. cu_seq_lens_k: (batch_size + 1,), dtype torch.int32. The cumulative sequence lengths of the sequences in the batch, used to index into kv. max_seqlen_q: int. Maximum query sequence length in the batch. max_seqlen_k: int. Maximum key sequence length in the batch. dropout_p: float. Dropout probability. softmax_scale: float. The scaling of QK^T before applying softmax. Default to 1 / sqrt(headdim). causal: bool. Whether to apply causal attention mask (e.g., for auto-regressive modeling). window_size: (left, right). If not (-1, -1), implements sliding window local attention. softcap: float. Anything > 0 activates softcapping attention. sliding_windowF)r rfull_attentionsliding_attentionN read_index write_index) key_states value_states layer_idxrrs_auxrT) softmax_scalecausal window_size)rconfig_attn_implementationgetattrr updater isinstancedictget transposesqueeze contiguoustotorchint32clonescalingtuple)rrrrr r cu_seq_lens_q cu_seq_lens_k max_seqlen_q max_seqlen_kkwargsflash_attn_varlen_funcr layer_type custom_kwargs attn_outputr3g/sda-disk/www/egybert/egybert_env/lib/python3.10/site-packages/transformers/integrations/flash_paged.pypaged_attention_forwardsB #    r5)NNNNNN) r%generation.continuous_batchingrmodeling_flash_attention_utilsrnnModuleTensorr5r3r3r3r4s0