feat: add act_quant_mode dropdown to ModelLoader and ModelAutoLoader
Allows selecting activation quantization scale algorithm: - absmax (default): fast, scale = absmax/7 - mse: search ~5 candidates for min MSE (+1dB quality, ~8% slower) Passed via config_json at pipeline creation time. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Y
yejianhui committed
82da33c75af1fb18309830776809e718d44e98c6
Parent: 2ac0bf3