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

fix clamp_kqv setting

Browse source
This commit is contained in:
nopperl 2024-04-18 23:39:18 +02:00
parent cc32c73926
commit f6c99ce011
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
convert-hf-to-gguf.py
View file

@ -2644,8 +2644,8 @@ class OlmoModel(Model):
def set_gguf_parameters(self): def set_gguf_parameters(self):
super().set_gguf_parameters() super().set_gguf_parameters()
self.gguf_writer.add_layer_norm_eps(1e-5) self.gguf_writer.add_layer_norm_eps(1e-5)
if self.hparams.get("clip_qkv") is not None: if "clip_qkv" in self.hparams is not None:
self.gguf_writer.add_clamp_kqv(self.hparams["attn_config"]["clip_qkv"]) self.gguf_writer.add_clamp_kqv(self.hparams["clip_qkv"])
# Same as super class, but permuting q_proj, k_proj # Same as super class, but permuting q_proj, k_proj
# Copied from: LlamaModel # Copied from: LlamaModel