root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-27 20:04:35 +00:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

gguf.py : fix rope scale kv

Browse Source
This commit is contained in:
klosax 2023-08-21 13:23:10 +02:00 committed by GitHub
parent c818c405e0
commit 6a69a693cb
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
gguf.py
View File

@ -45,7 +45,7 @@ KEY_ATTENTION_LAYERNORM_RMS_EPS = "{arch}.attention.layer_norm_rms_epsilon"
# RoPE
KEY_ROPE_DIMENSION_COUNT = "{arch}.rope.dimension_count"
KEY_ROPE_SCALE = "{arch}.rope.scale"
KEY_ROPE_SCALE_LINEAR = "{arch}.rope.scale_linear"
# tokenization
KEY_TOKENIZER_MODEL = "tokenizer.ggml.model"
@ -620,8 +620,8 @@ class GGUFWriter:
self.add_uint32(
KEY_ROPE_DIMENSION_COUNT.format(arch=self.arch), count)
def add_rope_scale(self, value: float):
self.add_float32(KEY_ROPE_SCALE.format(arch=self.arch), value)
def add_rope_scale_linear(self, value: float):
self.add_float32(KEY_ROPE_SCALE_LINEAR.format(arch=self.arch), value)
def add_tokenizer_model(self, model: str):
self.add_string(KEY_TOKENIZER_MODEL, model)