char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Xuan-Son Nguyenconvert: add @ModelBase.example (#27208)fa88ae936

char/aliceai-t5-moe
brancheschar/aliceai-t5-moeall refs →
837 B25 linesraw
1from __future__ import annotations
2
3from .base import ModelBase, gguf, logger
4from .llama import LlamaModel
5
6
7@ModelBase.register("NanbeigeForCausalLM")
8@ModelBase.example("Nanbeige/Nanbeige4.2-3B")
9class NanbeigeModel(LlamaModel):
10    model_arch = gguf.MODEL_ARCH.NANBEIGE
11    undo_permute = True
12
13    def set_gguf_parameters(self):
14        super().set_gguf_parameters()
15        hparams = self.hparams
16
17        n_loops = int(hparams.get("num_loops", 1) or 1)
18        if n_loops < 1:
19            n_loops = 1
20        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}")