Mojo function
mha_single_batch_gfx942
mha_single_batch_gfx942[output_type: DType, q_type: DType, k_t: MHAOperand, v_t: MHAOperand, mask_t: MHAMask, group: Int, config: MHAConfig, sink: Bool = False, sink_type: DType = output_type](output: UnsafePointer[Scalar[output_type]], q: UnsafePointer[Scalar[q_type]], k: k_t, v: v_t, seq_len: Int, num_keys: Int, scale: Float32, batch_idx: Int, start_pos: Int, mask: mask_t, sink_weights: OptionalReg[LayoutTensor[q_type, Layout.row_major(-1), MutableAnyOrigin]])
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!