CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : add jinja template for rwkv-world (#14665)

Browse Source 
* llama : add jinja template for rwkv-world

Signed-off-by: Molly Sophia <mollysophia379@gmail.com>

* Update convert_hf_to_gguf.py

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

---------

Signed-off-by: Molly Sophia <mollysophia379@gmail.com>
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
This commit is contained in:
Molly Sophia
2025-07-14 07:43:43 +08:00
committed by GitHub
parent 982e347255
commit 0d9226763c
3 changed files with 43 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
convert_hf_to_gguf.py
View File

@@ -1082,7 +1082,14 @@ class TextModel(ModelBase):
self.gguf_writer.add_token_list(tokens)
self.gguf_writer.add_token_types(toktypes)
special_vocab = gguf.SpecialVocab(self.dir_model, load_merges=False)
special_vocab.chat_template = "rwkv-world"
if special_vocab.chat_template is None:
template_path = Path(__file__).parent / "models" / "templates" / "llama-cpp-rwkv-world.jinja"
if template_path.is_file():
with open(template_path, "r", encoding="utf-8") as f:
template = f.read()
else:
template = "rwkv-world"
special_vocab.chat_template = template
# hack: Add '\n\n' as the EOT token to make it chat normally
special_vocab._set_special_token("eot", 261)
# hack: Override these as they have already been set (incorrectly)

34
models/templates/llama-cpp-rwkv-world.jinja Normal file
View File

@@ -0,0 +1,34 @@
{%- if not add_generation_prompt is defined -%}
{%- set add_generation_prompt = true -%}
{%- endif -%}
{%- set ns = namespace(system_prompt='') -%}
{%- for message in messages -%}
{%- if message['role'] == 'system' -%}
{%- set ns.system_prompt = message['content'] -%}
{%- endif -%}
{%- endfor -%}
{{bos_token}}
{%- if ns.system_prompt != '' -%}
{{- 'System: ' + ns.system_prompt + '\n\n' -}}
{%- endif -%}
{%- for message in messages -%}
{%- if message['role'] == 'user' -%}
{{- 'User: ' + message['content']|trim + '\n\n' -}}
{%- endif -%}
{%- if message['role'] == 'assistant' and message['content'] is not none -%}
{%- set content = message['content'] -%}
{%- if '</think>' in content -%}
{%- set content = content.split('</think>')[-1] -%}
{%- endif -%}
{{- 'Assistant: ' + content|trim + '\n\n' -}}
{%- endif -%}
{%- endfor -%}
{%- if add_generation_prompt -%}
{{- 'Assistant:' -}}
{%- if enable_thinking is defined and enable_thinking is false %}
{{- ' <think>\n</think>' }}
{%- endif %}
{%- if enable_thinking is defined and enable_thinking is true %}
{{- ' <think>' }}
{%- endif %}
{%- endif -%}

2
src/llama-chat.cpp
View File

@@ -170,7 +170,7 @@ llm_chat_template llm_chat_detect_template(const std::string & tmpl) {
// ref: https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct/discussions/8#66bae61b1893d14ee8ed85bb
// EXAONE-3.0-7.8B-Instruct
return LLM_CHAT_TEMPLATE_EXAONE_3;
} else if (tmpl_contains("rwkv-world")) {
} else if (tmpl_contains("rwkv-world") || tmpl_contains("{{- 'User: ' + message['content']|trim + '\\n\\n' -}}")) {
return LLM_CHAT_TEMPLATE_RWKV_WORLD;
} else if (tmpl_contains("<|start_of_role|>")) {
return LLM_CHAT_TEMPLATE_GRANITE;