flashinfer.comm.mnnvl_cutedsl_ag.MNNVLCuteDSLAllGatherDualNormWorkspace

class flashinfer.comm.mnnvl_cutedsl_ag.MNNVLCuteDSLAllGatherDualNormWorkspace(tp_size: int, tp_rank: int, max_token_num: int, hidden_dim: int, *, group: ProcessGroup, tuning: LLAllGatherDualNormTuning | None = None)

__init__(tp_size: int, tp_rank: int, max_token_num: int, hidden_dim: int, *, group: ProcessGroup, tuning: LLAllGatherDualNormTuning | None = None) → None

Methods

__init__(tp_size, tp_rank, max_token_num, ...)
check_support(local_supported)	Collectively agree on a local supports_* result before capture.
compile(*, post_eps, next_eps[, ...])	Compile one constant set ahead of time, e.g. before graph capture.
destroy()	Release the mailbox once in-flight work on every rank is done.
fits(shard)	Whether this rank's [M, H/TP] BF16 shard fits the workspace.

Attributes

impl