mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2026-08-27 07:31:24 +02:00
Xuan-Son Nguyen
fa88ae9368
* convert: add @ModelBase.example * add docs * add more variants * BailingMoeV3ForCausalLM * rm pocket-tts
26 lines
837 B
Python
26 lines
837 B
Python
from __future__ import annotations
|
|
|
|
from .base import ModelBase, gguf, logger
|
|
from .llama import LlamaModel
|
|
|
|
|
|
@ModelBase.register("NanbeigeForCausalLM")
|
|
@ModelBase.example("Nanbeige/Nanbeige4.2-3B")
|
|
class NanbeigeModel(LlamaModel):
|
|
model_arch = gguf.MODEL_ARCH.NANBEIGE
|
|
undo_permute = True
|
|
|
|
def set_gguf_parameters(self):
|
|
super().set_gguf_parameters()
|
|
hparams = self.hparams
|
|
|
|
n_loops = int(hparams.get("num_loops", 1) or 1)
|
|
if n_loops < 1:
|
|
n_loops = 1
|
|
self.gguf_writer.add_num_loops(n_loops)
|
|
logger.info(f"gguf: num_loops = {n_loops}")
|
|
|
|
skip_loop_final_norm = bool(hparams.get("skip_loop_final_norm", False))
|
|
self.gguf_writer.add_skip_loop_final_norm(skip_loop_final_norm)
|
|
logger.info(f"gguf: skip_loop_final_norm = {skip_loop_final_norm}")
|