o ?߱i@sdZddlmZddlmZddlmZmZddlm Z ddl m Z m Z ddl mZddlmZmZ dd ed ed ed ededededefddZdS)z Imaginaire4 Attention Subpackage: Unified implementation for all Attention implementations. Flash Attention v2 (flash2) backend checks )partial)Tensor)attention_param_checksattention_tensor_checks)FLASH2_SUPPORTED)get_bwd_dtypesget_fwd_dtypes) CausalType) get_arch_taglog_or_raise_errorFquerykeyvalue is_causal causal_type is_varlen raise_errorreturnc Cstt|d}ts|dtddS|r|dtddSt|j}t|} t|} t |||| | dd|dd s<|d tddSt |||||d |rW|t j t j fvrW|d tddSdS) a Input validation function for the flash2 backend. Parameters: query (Tensor): 4-D query tensor, with the heads-last contiguous layout (`[batch, seqlen, heads, head_dim]`). key (Tensor): 4-D key tensor, with the heads-last contiguous layout (`[batch, seqlen_kv, heads_kv, head_dim]`). value (Tensor): 4-D value tensor, with heads-last contiguous layout (`[batch, seqlen_kv, heads_kv, head_dim_v]`). is_causal (bool): whether or not causal masking is enabled. causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, `CausalType.BottomRight`. Required when `is_causal = True`. is_varlen (bool): whether or not a variable length (varlen) use case. Must be inferred beforehand based on arguments such as seqlens_{Q,KV} or cumulative_seqlen_{Q,KV} being passed. raise_error (bool): whether to raise an error if any checks fail or no backend is selected, instead of just returning False. Default is False. Returns: success (bool): whether use case is compatible with flash2 backend. )rzFlash Attention v2 (flash2) is not supported in this environment. Run with debug logs to find out why, or choose another backend.) exceptionFz_Flash Attention v2 (flash2) varlen is banned due to instability. Please choose another backend.TzFlash Attention v2 (flash2)) r r rZsupported_dtypes_forwardZsupported_dtypes_backwardZ supports_mlaZsupports_gqa_mqar backend_namez>Flash Attention v2 (flash2) does not support the given inputs.)r r rrrz=Flash Attention v2 only supports bottom-right causal masking.)rr r RuntimeError ValueErrorr devicerrrrr Z BottomRightDontCare) r r rrrrr target_fnarch_tagZ fwd_dtypesZ bwd_dtypesra/data/cameron/vidgen/cosmos-predict2.5/cosmos_predict2/_src/imaginaire/attention/flash2/checks.pyflash2_attention_check"sP &   rN)F)__doc__ functoolsrtorchr0cosmos_predict2._src.imaginaire.attention.checksrr0cosmos_predict2._src.imaginaire.attention.flash2rZ5cosmos_predict2._src.imaginaire.attention.flash2.metarr/cosmos_predict2._src.imaginaire.attention.masksr /cosmos_predict2._src.imaginaire.attention.utilsr r boolrrrrrs4