Felipe97/llama-cpp-compiled

1from __future__ import annotations2 3from .base import ModelBase, gguf, logger4from .llama import LlamaModel5 6 7@ModelBase.register("NanbeigeForCausalLM")8@ModelBase.example("Nanbeige/Nanbeige4.2-3B")9class NanbeigeModel(LlamaModel):10    model_arch = gguf.MODEL_ARCH.NANBEIGE11    undo_permute = True12 13    def set_gguf_parameters(self):14        super().set_gguf_parameters()15        hparams = self.hparams16 17        n_loops = int(hparams.get("num_loops", 1) or 1)18        if n_loops < 1:19            n_loops = 120        self.gguf_writer.add_num_loops(n_loops)21        logger.info(f"gguf: num_loops = {n_loops}")22 23        skip_loop_final_norm = bool(hparams.get("skip_loop_final_norm", False))24        self.gguf_writer.add_skip_loop_final_norm(skip_loop_final_norm)25        logger.info(f"gguf: skip_loop_final_norm = {skip_loop_final_norm}")26