CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

gguf : add rope_freq_base parameter for CodeLlama (#2769)

Browse Source
This commit is contained in:
slaren
2023-08-24 20:04:05 +02:00
committed by GitHub
parent 01f2224682
commit 0d3094f0c7
3 changed files with 48 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
gguf.py
View File

@@ -47,6 +47,7 @@ KEY_ATTENTION_LAYERNORM_RMS_EPS = "{arch}.attention.layer_norm_rms_epsilon"
# RoPE
KEY_ROPE_DIMENSION_COUNT = "{arch}.rope.dimension_count"
KEY_ROPE_FREQ_BASE = "{arch}.rope.freq_base"
KEY_ROPE_SCALE_LINEAR = "{arch}.rope.scale_linear"
# tokenization
@@ -663,7 +664,10 @@ class GGUFWriter:
self.add_uint32(
KEY_ROPE_DIMENSION_COUNT.format(arch=self.arch), count)
def add_rope_scale_linear(self, value: float):
def add_rope_freq_base(self, value: float):
self.add_float32(KEY_ROPE_FREQ_BASE.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):