Expand description
Mitra inference on the Burn backend (Flex CPU).
Quantile-bucketize embedding, row/feature dual attention blocks, gated
head readout, with weights loaded straight from GGUF via
zsfm_burn::GgufWeights.
Mitra inference on the Burn backend (Flex CPU).
Quantile-bucketize embedding, row/feature dual attention blocks, gated
head readout, with weights loaded straight from GGUF via
zsfm_burn::GgufWeights.