Ratio controlling the rotary cache magnitude

Read the equation

This coefficient divides the adjustment computed with mscale by the adjustment computed with mscale_all_dim. Equal strengths give a ratio of one. The attention initializer can still multiply its own coefficient by a squared adjustment, so an unchanged rotary-cache magnitude does not imply unchanged attention scaling.

Terms & definitions

The recovered passage does not supply a separate symbol table. Read the source context for the notation used in this version; familiar letters may carry different meanings in other builds.

How it works

Read the definition from its right-hand expression to the quantity on the left. Its surrounding source records the intended input, state, or parameter context. Where the source leaves an operation unspecified, this page preserves that gap.

Assumptions & limits

Source recovery establishes that the expression appears in this record. Its mathematical validity, implementation behavior, and interpretation must be assessed separately. See the linked source and research notes for the checks available for this version.

Original context

freqs = torch.outer(t, inv_freq)

        _mscale = float(
            yarn_get_mscale(self.scaling_factor, self.mscale) /
            yarn_get_mscale(self.scaling_factor, self.mscale_all_dim))

        emb = torch.cat((freqs, freqs), dim=-1)
        self.register_buffer("cos_cached", (emb.cos() * _mscale).to(dtype),

_mscale = float(
            yarn_get_mscale(self.scaling_factor, self.mscale) /
            yarn_get_mscale(self.scaling_factor, self.mscale_all_dim))

Related builds