• Nicolas Patry's avatar
    Tied embeddings in MLP speculator. (#2473) · d9fbbaaf
    Nicolas Patry authored
    * Tied embeddings in MLP speculator.
    
    * Fixing the scale_weight when users decide to not use the speculation as
    much as defined in the config.
    
    * Adding scaling support + optimize some ops.
    d9fbbaaf
mlp.py 9.89 KB