CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-02 09:12:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into compilade/refactor-kv-cache

Browse Source
This commit is contained in:
Francis Couture-Harpin
2025-07-03 16:03:56 -04:00
parent 908e6559d6 bee28421be
commit 4682e21c46
89 changed files with 895 additions and 4844 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
gguf-py/gguf/gguf_writer.py
View File

@@ -714,8 +714,8 @@ class GGUFWriter:
def add_clamp_kqv(self, value: float) -> None:
self.add_float32(Keys.Attention.CLAMP_KQV.format(arch=self.arch), value)
def add_shared_kv_layers(self, value: float) -> None:
self.add_float32(Keys.Attention.SHARED_KV_LAYERS.format(arch=self.arch), value)
def add_shared_kv_layers(self, value: int) -> None:
self.add_uint32(Keys.Attention.SHARED_KV_LAYERS.format(arch=self.arch), value)
def add_sliding_window_pattern(self, value: Sequence[bool]) -> None:
self.add_array(Keys.Attention.SLIDING_WINDOW_PATTERN.format(arch=self.arch), value)

15
gguf-py/gguf/vocab.py
View File

@@ -245,9 +245,18 @@ class SpecialVocab:
if not tokenizer_config:
return True
chat_template_alt = None
chat_template_file = path / 'chat_template.json'
if chat_template_file.is_file():
with open(chat_template_file, encoding = 'utf-8') as f:
chat_template_json = path / 'chat_template.json'
chat_template_jinja = path / 'chat_template.jinja'
if chat_template_jinja.is_file():
with open(chat_template_jinja, encoding = 'utf-8') as f:
chat_template_alt = f.read()
if additional_templates := list((path / 'additional_chat_templates').glob('*.jinja')):
chat_template_alt = [{'name': 'default', 'template': chat_template_alt}]
for template_path in additional_templates:
with open(template_path, encoding = 'utf-8') as fp:
chat_template_alt.append({'name': template_path.stem, 'template': fp.read()})
elif chat_template_json.is_file():
with open(chat_template_json, encoding = 'utf-8') as f:
chat_template_alt = json.load(f).get('chat_template')
chat_template = tokenizer_config.get('chat_template', chat_template_alt)
if chat_template is None or isinstance(chat_template, (str, list)):