vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Yarn rope for baichuan

Browse source
This commit is contained in:
Galunid 2023-11-04 23:15:41 +01:00
parent e64f4de189
commit 2120195bb1
1 changed files with 2 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

3
model.py
View file

@ -505,7 +505,8 @@ class BaichuanModel(Model):
if self.hparams.get("rope_scaling") is not None and "factor" in self.hparams["rope_scaling"]: if self.hparams.get("rope_scaling") is not None and "factor" in self.hparams["rope_scaling"]:
if self.hparams["rope_scaling"].get("type") == "linear": if self.hparams["rope_scaling"].get("type") == "linear":
self.gguf_writer.add_rope_scale_linear(self.hparams["rope_scaling"]["factor"]) self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.LINEAR)
self.gguf_writer.add_rope_scaling_factor(self.hparams["rope_scaling"]["factor"])
def write_tensors(self): def write_tensors(self):
# Collect tensors from generator object # Collect tensors from generator object