CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-14 11:07:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into compilade/mamba2

Browse Source
This commit is contained in:
Francis Couture-Harpin
2025-02-26 15:32:20 -05:00
parent 1ee6c482d0 a800ae46da
commit c9ecf620d6
702 changed files with 98035 additions and 62199 deletions
Download Patch File Download Diff File Expand all files Collapse all files

72
tests/CMakeLists.txt
View File

@@ -1,3 +1,5 @@
llama_add_compile_flags()
function(llama_test target)
include(CMakeParseArguments)
set(options)
@@ -84,39 +86,56 @@ llama_test(test-tokenizer-0 NAME test-tokenizer-0-qwen2 ARGS ${CMAKE
llama_test(test-tokenizer-0 NAME test-tokenizer-0-refact ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-refact.gguf)
llama_test(test-tokenizer-0 NAME test-tokenizer-0-starcoder ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-starcoder.gguf)
# build test-tokenizer-1-bpe target once and add many tests
add_executable(test-tokenizer-1-bpe test-tokenizer-1-bpe.cpp)
target_link_libraries(test-tokenizer-1-bpe PRIVATE common)
install(TARGETS test-tokenizer-1-bpe RUNTIME)
if (LLAMA_LLGUIDANCE)
llama_target_and_test(test-grammar-llguidance.cpp ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-llama-bpe.gguf)
endif ()
# TODO: disabled due to slowness
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-aquila ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-aquila.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-falcon ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-falcon.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-gpt-2 ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-gpt-2.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-gpt-neox ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-gpt-neox.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-llama-bpe ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-llama-bpe.gguf --ignore-merges)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-mpt ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-mpt.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-refact ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-refact.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-starcoder ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-starcoder.gguf)
if (NOT WIN32)
# these tests are disabled on Windows because they use internal functions not exported with LLAMA_API
llama_target_and_test(test-sampling.cpp)
llama_target_and_test(test-grammar-parser.cpp)
llama_target_and_test(test-grammar-integration.cpp)
llama_target_and_test(test-llama-grammar.cpp)
llama_target_and_test(test-chat.cpp)
# TODO: disabled on loongarch64 because the ggml-ci node lacks Python 3.8
if (NOT ${CMAKE_SYSTEM_PROCESSOR} MATCHES "loongarch64")
llama_target_and_test(test-json-schema-to-grammar.cpp WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/..)
target_include_directories(test-json-schema-to-grammar PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/../examples/server)
endif()
# build test-tokenizer-1-spm target once and add many tests
add_executable(test-tokenizer-1-spm test-tokenizer-1-spm.cpp)
target_link_libraries(test-tokenizer-1-spm PRIVATE common)
install(TARGETS test-tokenizer-1-spm RUNTIME)
llama_test(test-tokenizer-1-spm NAME test-tokenizer-1-llama-spm ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-llama-spm.gguf)
#llama_test(test-tokenizer-1-spm NAME test-tokenizer-1-baichuan ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-baichuan.gguf)
# build test-tokenizer-1-bpe target once and add many tests
add_executable(test-tokenizer-1-bpe test-tokenizer-1-bpe.cpp)
target_link_libraries(test-tokenizer-1-bpe PRIVATE common)
install(TARGETS test-tokenizer-1-bpe RUNTIME)
# TODO: disabled due to slowness
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-aquila ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-aquila.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-falcon ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-falcon.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-gpt-2 ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-gpt-2.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-gpt-neox ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-gpt-neox.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-llama-bpe ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-llama-bpe.gguf --ignore-merges)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-mpt ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-mpt.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-refact ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-refact.gguf)
#llama_test(test-tokenizer-1-bpe NAME test-tokenizer-1-starcoder ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-starcoder.gguf)
# build test-tokenizer-1-spm target once and add many tests
add_executable(test-tokenizer-1-spm test-tokenizer-1-spm.cpp)
target_link_libraries(test-tokenizer-1-spm PRIVATE common)
install(TARGETS test-tokenizer-1-spm RUNTIME)
llama_test(test-tokenizer-1-spm NAME test-tokenizer-1-llama-spm ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-llama-spm.gguf)
#llama_test(test-tokenizer-1-spm NAME test-tokenizer-1-baichuan ARGS ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-baichuan.gguf)
# llama_target_and_test(test-double-float.cpp) # SLOW
endif()
# llama_target_and_test(test-double-float.cpp) # SLOW
llama_target_and_test(test-log.cpp)
llama_target_and_test(test-arg-parser.cpp)
llama_target_and_test(test-sampling.cpp)
llama_target_and_test(test-chat-template.cpp)
llama_target_and_test(test-grammar-parser.cpp)
llama_target_and_test(test-grammar-integration.cpp)
llama_target_and_test(test-llama-grammar.cpp)
# llama_target_and_test(test-opt.cpp) # SLOW
llama_target_and_test(test-gguf.cpp)
llama_target_and_test(test-backend-ops.cpp)
llama_target_and_test(test-model-load-cancel.cpp LABEL "model")
@@ -130,11 +149,6 @@ if (NOT GGML_BACKEND_DL)
llama_target_and_test(test-rope.cpp)
endif()
# TODO: disabled on loongarch64 because the ggml-ci node lacks Python 3.8
if (NOT ${CMAKE_SYSTEM_PROCESSOR} MATCHES "loongarch64")
llama_target_and_test(test-json-schema-to-grammar.cpp WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/..)
target_include_directories(test-json-schema-to-grammar PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/../examples/server)
endif()
# dummy executable - not installed
get_filename_component(TEST_TARGET test-c.c NAME_WE)

6
tests/test-autorelease.cpp
View File

@@ -13,10 +13,10 @@ int main(int argc, char ** argv) {
std::thread([&model_path]() {
llama_backend_init();
auto * model = llama_load_model_from_file(model_path, llama_model_default_params());
auto * ctx = llama_new_context_with_model(model, llama_context_default_params());
auto * model = llama_model_load_from_file(model_path, llama_model_default_params());
auto * ctx = llama_init_from_model(model, llama_context_default_params());
llama_free(ctx);
llama_free_model(model);
llama_model_free(model);
llama_backend_free();
}).join();

728
tests/test-backend-ops.cpp
View File

File diff suppressed because it is too large Load Diff

458
tests/test-chat-template.cpp
View File

@@ -1,15 +1,33 @@
#include <string>
#include <vector>
#include <sstream>
#include <regex>
#undef NDEBUG
#include <cassert>
#include "llama.h"
#include "common.h"
#include "chat.h"
static std::string normalize_newlines(const std::string & s) {
#ifdef _WIN32
static const std::regex nl_regex("\r\n");
return std::regex_replace(s, nl_regex, "\n");
#else
return s;
#endif
}
static common_chat_msg simple_msg(const std::string & role, const std::string & content) {
common_chat_msg msg;
msg.role = role;
msg.content = content;
return msg;
}
int main(void) {
llama_chat_message conversation[] = {
std::vector<llama_chat_message> conversation {
{"system", "You are a helpful assistant"},
{"user", "Hello"},
{"assistant", "Hi there"},
@@ -17,165 +35,363 @@ int main(void) {
{"assistant", " I am an assistant "},
{"user", "Another question"},
};
size_t message_count = 6;
std::vector<std::string> templates = {
// teknium/OpenHermes-2.5-Mistral-7B
"{% for message in messages %}{{'<|im_start|>' + message['role'] + '\\n' + message['content'] + '<|im_end|>' + '\\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\\n' }}{% endif %}",
// mistralai/Mistral-7B-Instruct-v0.2
"{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}",
// TheBloke/FusionNet_34Bx2_MoE-AWQ
"{%- for idx in range(0, messages|length) -%}\\n{%- if messages[idx]['role'] == 'user' -%}\\n{%- if idx > 1 -%}\\n{{- bos_token + '[INST] ' + messages[idx]['content'] + ' [/INST]' -}}\\n{%- else -%}\\n{{- messages[idx]['content'] + ' [/INST]' -}}\\n{%- endif -%}\\n{% elif messages[idx]['role'] == 'system' %}\\n{{- '[INST] <<SYS>>\\\\n' + messages[idx]['content'] + '\\\\n<</SYS>>\\\\n\\\\n' -}}\\n{%- elif messages[idx]['role'] == 'assistant' -%}\\n{{- ' ' + messages[idx]['content'] + ' ' + eos_token -}}\\n{% endif %}\\n{% endfor %}",
// bofenghuang/vigogne-2-70b-chat
"{{ bos_token }}{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% elif true == true and not '<<SYS>>' in messages[0]['content'] %}{% set loop_messages = messages %}{% set system_message = 'Vous êtes Vigogne, un assistant IA créé par Zaion Lab. Vous suivez extrêmement bien les instructions. Aidez autant que vous le pouvez.' %}{% else %}{% set loop_messages = messages %}{% set system_message = false %}{% endif %}{% for message in loop_messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if loop.index0 == 0 and system_message != false %}{% set content = '<<SYS>>\\\\n' + system_message + '\\\\n<</SYS>>\\\\n\\\\n' + message['content'] %}{% else %}{% set content = message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + content.strip() + ' [/INST]' }}{% elif message['role'] == 'system' %}{{ '<<SYS>>\\\\n' + content.strip() + '\\\\n<</SYS>>\\\\n\\\\n' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content.strip() + ' ' + eos_token }}{% endif %}{% endfor %}",
// mlabonne/AlphaMonarch-7B
"{% for message in messages %}{{bos_token + message['role'] + '\\n' + message['content'] + eos_token + '\\n'}}{% endfor %}{% if add_generation_prompt %}{{ bos_token + 'assistant\\n' }}{% endif %}",
// google/gemma-7b-it
"{% if messages[0]['role'] == 'system' %}{{ raise_exception('System role not supported') }}{% endif %}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if (message['role'] == 'assistant') %}{% set role = 'model' %}{% else %}{% set role = message['role'] %}{% endif %}{{ '<start_of_turn>' + role + '\\n' + message['content'] | trim + '<end_of_turn>\\n' }}{% endfor %}{% if add_generation_prompt %}{{'<start_of_turn>model\\n'}}{% endif %}",
// OrionStarAI/Orion-14B-Chat
"{% for message in messages %}{% if loop.first %}{{ bos_token }}{% endif %}{% if message['role'] == 'user' %}{{ 'Human: ' + message['content'] + '\\n\\nAssistant: ' + eos_token }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token }}{% endif %}{% endfor %}",
// openchat/openchat-3.5-0106
// The included chat_template differs from the author's suggestions here: https://huggingface.co/openchat/openchat_3.5/discussions/5#65448109b4a3f3a2f486fd9d
// So we match against the included template but implement the suggested version.
"{{ bos_token }}{% for message in messages %}{{ 'GPT4 Correct ' + message['role'].title() + ': ' + message['content'] + '<|end_of_turn|>'}}{% endfor %}{% if add_generation_prompt %}{{ 'GPT4 Correct Assistant:' }}{% endif %}",
// deepseek-ai/deepseek-coder-33b-instruct
"{% if not add_generation_prompt is defined %}\n{% set add_generation_prompt = false %}\n{% endif %}\n{%- set ns = namespace(found=false) -%}\n{%- for message in messages -%}\n {%- if message['role'] == 'system' -%}\n {%- set ns.found = true -%}\n {%- endif -%}\n{%- endfor -%}\n{{bos_token}}{%- if not ns.found -%}\n{{'You are an AI programming assistant, utilizing the Deepseek Coder model, developed by Deepseek Company, and you only answer questions related to computer science. For politically sensitive questions, security and privacy issues, and other non-computer science questions, you will refuse to answer\\n'}}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'system' %}\n{{ message['content'] }}\n {%- else %}\n {%- if message['role'] == 'user' %}\n{{'### Instruction:\\n' + message['content'] + '\\n'}}\n {%- else %}\n{{'### Response:\\n' + message['content'] + '\\n<|EOT|>\\n'}}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{% if add_generation_prompt %}\n{{'### Response:'}}\n{% endif %}",
// eachadea/vicuna-13b-1.1
// No template included in tokenizer_config.json, so this template likely needs to be manually set.
"{%- for message in messages %}{%- if message['role'] == 'system' -%}{{- '' + message['content'] + '\n\n' -}}{%- else -%}{%- if message['role'] == 'user' -%}{{-'USER: ' + message['content'] + '\n'-}}{%- else -%}{{-'ASSISTANT: ' + message['content'] + '</s>\n' -}}{%- endif -%}{%- endif -%}{%- endfor -%}{%- if add_generation_prompt -%}{{-'ASSISTANT:'-}}{%- endif -%}",
// Orca-Vicuna
// No template included in tokenizer_config.json, so this template likely needs to be manually set.
"{%- for message in messages %}{%- if message['role'] == 'system' -%}{{-'SYSTEM: ' + message['content'] + '\n' -}}{%- else -%}{%- if message['role'] == 'user' -%}{{-'USER: ' + message['content'] + '\n'-}}{%- else -%}{{-'ASSISTANT: ' + message['content'] + '</s>\n' -}}{%- endif -%}{%- endif -%}{%- endfor -%}{%- if add_generation_prompt -%}{{-'ASSISTANT:'-}}{%- endif -%}",
// CohereForAI/c4ai-command-r-plus
"{{ bos_token }}{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% elif false == true %}{% set loop_messages = messages %}{% set system_message = 'You are Command-R, a brilliant, sophisticated, AI-assistant trained to assist human users by providing thorough responses. You are trained by Cohere.' %}{% else %}{% set loop_messages = messages %}{% set system_message = false %}{% endif %}{% if system_message != false %}{{ '<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>' + system_message + '<|END_OF_TURN_TOKEN|>' }}{% endif %}{% for message in loop_messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<|START_OF_TURN_TOKEN|><|USER_TOKEN|>' + content.strip() + '<|END_OF_TURN_TOKEN|>' }}{% elif message['role'] == 'assistant' %}{{ '<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>' + content.strip() + '<|END_OF_TURN_TOKEN|>' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>' }}{% endif %}",
// Llama-3
"{% set loop_messages = messages %}{% for message in loop_messages %}{% set content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' %}{% if loop.index0 == 0 %}{% set content = bos_token + content %}{% endif %}{{ content }}{% endfor %}{{ '<|start_header_id|>assistant<|end_header_id|>\n\n' }}",
//Phi-3-mini
"{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') %}{{'<|user|>' + '\n' + message['content'] + '<|end|>' + '\n' + '<|assistant|>' + '\n'}}{% elif (message['role'] == 'assistant') %}{{message['content'] + '<|end|>' + '\n'}}{% endif %}{% endfor %}",
//Phi-3-small
"{{ bos_token }}{% for message in messages %}{{'<|' + message['role'] + '|>' + '\n' + message['content'] + '<|end|>\n' }}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>\n' }}{% else %}{{ eos_token }}{% endif %}",
//Phi-3-medium
"{% for message in messages %}{% if (message['role'] == 'user') %}{{'<|user|>' + '\n' + message['content'] + '<|end|>' + '\n' + '<|assistant|>' + '\n'}}{% elif (message['role'] == 'assistant') %}{{message['content'] + '<|end|>' + '\n'}}{% endif %}{% endfor %}",
//Phi-3-vision
"{% for message in messages %}{{'<|' + message['role'] + '|>' + '\n' + message['content'] + '<|end|>\n' }}{% endfor %}{% if add_generation_prompt and messages[-1]['role'] != 'assistant' %}{{- '<|assistant|>\n' -}}{% endif %}",
// ChatGLM3
"{% for message in messages %}{% if loop.first %}[gMASK]sop<|{{ message['role'] }}|>\n {{ message['content'] }}{% else %}<|{{ message['role'] }}|>\n {{ message['content'] }}{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>{% endif %}",
// ChatGLM4
u8"[gMASK]<sop>{% for item in messages %}{% if item['tools'] is defined %}<|system|>\n你是一个名为 ChatGLM 的人工智能助手。你是基于智谱AI训练的语言模型 GLM-4 模型开发的，你的任务是针对用户的问题和要求提供适当的答复和支持。\n\n# 可用工具{% set tools = item['tools'] %}{% for tool in tools %}{% if tool['type'] == 'function' %}\n\n## {{ tool['function']['name'] }}\n\n{{ tool['function'] | tojson(indent=4) }}\n......{% endif %}{% endfor %}{% endif %}{% if item['content'] %}<|{{ item['role'] }}|>{{ item['metadata'] }}\n{{ item['content'] }}{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>{% endif %}",
// MiniCPM-3B-OpenHermes-2.5-v2-GGUF
u8"{% for message in messages %}{% if message['role'] == 'user' %}{{'<用户>' + message['content'].strip() + '<AI>'}}{% else %}{{message['content'].strip()}}{% endif %}{% endfor %}",
// DeepSeek-V2
"{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ 'User: ' + message['content'] + '\n\n' }}{% elif message['role'] == 'assistant' %}{{ 'Assistant: ' + message['content'] + eos_token }}{% elif message['role'] == 'system' %}{{ message['content'] + '\n\n' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ 'Assistant:' }}{% endif %}",
// ibm-granite/granite-3.0-8b-instruct
"{%- if tools %}\n {{- '<|start_of_role|>available_tools<|end_of_role|>\n' }}\n {%- for tool in tools %}\n {{- tool | tojson(indent=4) }}\n {%- if not loop.last %}\n {{- '\n\n' }}\n {%- endif %}\n {%- endfor %}\n {{- '<|end_of_text|>\n' }}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'system' %}\n {{- '<|start_of_role|>system<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'user' %}\n {{- '<|start_of_role|>user<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'assistant' %}\n {{- '<|start_of_role|>assistant<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'assistant_tool_call' %}\n {{- '<|start_of_role|>assistant<|end_of_role|><|tool_call|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'tool_response' %}\n {{- '<|start_of_role|>tool_response<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- endif %}\n {%- if loop.last and add_generation_prompt %}\n {{- '<|start_of_role|>assistant<|end_of_role|>' }}\n {%- endif %}\n{%- endfor %}",
struct TestCase {
std::string name;
std::string template_str;
std::string expected_output;
std::string expected_output_jinja;
std::string bos_token = "";
std::string eos_token = "";
bool supported_with_jinja = true;
};
std::vector<std::string> expected_output = {
// teknium/OpenHermes-2.5-Mistral-7B
"<|im_start|>system\nYou are a helpful assistant<|im_end|>\n<|im_start|>user\nHello<|im_end|>\n<|im_start|>assistant\nHi there<|im_end|>\n<|im_start|>user\nWho are you<|im_end|>\n<|im_start|>assistant\n I am an assistant <|im_end|>\n<|im_start|>user\nAnother question<|im_end|>\n<|im_start|>assistant\n",
// mistralai/Mistral-7B-Instruct-v0.2
"[INST] You are a helpful assistant\nHello [/INST]Hi there</s>[INST] Who are you [/INST] I am an assistant </s>[INST] Another question [/INST]",
// TheBloke/FusionNet_34Bx2_MoE-AWQ
"[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST] Hi there </s><s>[INST] Who are you [/INST] I am an assistant </s><s>[INST] Another question [/INST]",
// bofenghuang/vigogne-2-70b-chat
"[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST] Hi there </s>[INST] Who are you [/INST] I am an assistant </s>[INST] Another question [/INST]",
// mlabonne/AlphaMonarch-7B
"system\nYou are a helpful assistant</s>\n<s>user\nHello</s>\n<s>assistant\nHi there</s>\n<s>user\nWho are you</s>\n<s>assistant\n I am an assistant </s>\n<s>user\nAnother question</s>\n<s>assistant\n",
// google/gemma-7b-it
"<start_of_turn>user\nYou are a helpful assistant\n\nHello<end_of_turn>\n<start_of_turn>model\nHi there<end_of_turn>\n<start_of_turn>user\nWho are you<end_of_turn>\n<start_of_turn>model\nI am an assistant<end_of_turn>\n<start_of_turn>user\nAnother question<end_of_turn>\n<start_of_turn>model\n",
// OrionStarAI/Orion-14B-Chat
"Human: You are a helpful assistant\n\nHello\n\nAssistant: </s>Hi there</s>Human: Who are you\n\nAssistant: </s> I am an assistant </s>Human: Another question\n\nAssistant: </s>",
// openchat/openchat-3.5-0106
"You are a helpful assistant<|end_of_turn|>GPT4 Correct User: Hello<|end_of_turn|>GPT4 Correct Assistant: Hi there<|end_of_turn|>GPT4 Correct User: Who are you<|end_of_turn|>GPT4 Correct Assistant: I am an assistant <|end_of_turn|>GPT4 Correct User: Another question<|end_of_turn|>GPT4 Correct Assistant:",
// deepseek-ai/deepseek-coder-33b-instruct
"You are a helpful assistant### Instruction:\nHello\n### Response:\nHi there\n<|EOT|>\n### Instruction:\nWho are you\n### Response:\n I am an assistant \n<|EOT|>\n### Instruction:\nAnother question\n### Response:\n",
// eachadea/vicuna-13b-1.1
"You are a helpful assistant\n\nUSER: Hello\nASSISTANT: Hi there</s>\nUSER: Who are you\nASSISTANT: I am an assistant </s>\nUSER: Another question\nASSISTANT:",
// Orca-Vicuna
"SYSTEM: You are a helpful assistant\nUSER: Hello\nASSISTANT: Hi there</s>\nUSER: Who are you\nASSISTANT: I am an assistant </s>\nUSER: Another question\nASSISTANT:",
// CohereForAI/c4ai-command-r-plus
"<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>You are a helpful assistant<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>Hi there<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Who are you<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>I am an assistant<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Another question<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>",
// Llama 3
"<|start_header_id|>system<|end_header_id|>\n\nYou are a helpful assistant<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nHello<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\nHi there<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nWho are you<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\nI am an assistant<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nAnother question<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
//Phi-3-mini
"<|system|>\nYou are a helpful assistant<|end|>\n<|user|>\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
//Phi-3-small
"<|system|>\nYou are a helpful assistant<|end|>\n<|user|>\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
//Phi-3-medium
"<|system|>\nYou are a helpful assistant<|end|>\n<|user|>\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
//Phi-3-vision
"<|system|>\nYou are a helpful assistant<|end|>\n<|user|>\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
// ChatGLM3
"[gMASK]sop<|system|>\n You are a helpful assistant<|user|>\n Hello<|assistant|>\n Hi there<|user|>\n Who are you<|assistant|>\n I am an assistant <|user|>\n Another question<|assistant|>",
// ChatGLM4
"[gMASK]<sop><|system|>\nYou are a helpful assistant<|user|>\nHello<|assistant|>\nHi there<|user|>\nWho are you<|assistant|>\n I am an assistant <|user|>\nAnother question<|assistant|>",
// MiniCPM-3B-OpenHermes-2.5-v2-GGUF
u8"You are a helpful assistant<用户>Hello<AI>Hi there<用户>Who are you<AI>I am an assistant<用户>Another question<AI>",
// DeepSeek-V2
u8"You are a helpful assistant\n\nUser: Hello\n\nAssistant: Hi there<｜end▁of▁sentence｜>User: Who are you\n\nAssistant: I am an assistant <｜end▁of▁sentence｜>User: Another question\n\nAssistant:",
// ibm-granite/granite-3.0-8b-instruct
"<|start_of_role|>system<|end_of_role|>You are a helpful assistant<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Hello<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>Hi there<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Who are you<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|> I am an assistant <|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Another question<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>\n",
std::vector<TestCase> test_cases {
{
/* .name= */ "teknium/OpenHermes-2.5-Mistral-7B",
/* .template_str= */ "{% for message in messages %}{{'<|im_start|>' + message['role'] + '\\n' + message['content'] + '<|im_end|>' + '\\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\\n' }}{% endif %}",
/* .expected_output= */ "<|im_start|>system\nYou are a helpful assistant<|im_end|>\n<|im_start|>user\nHello<|im_end|>\n<|im_start|>assistant\nHi there<|im_end|>\n<|im_start|>user\nWho are you<|im_end|>\n<|im_start|>assistant\n I am an assistant <|im_end|>\n<|im_start|>user\nAnother question<|im_end|>\n<|im_start|>assistant\n",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
{
/* .name= */ "mistralai/Mistral-7B-Instruct-v0.2 (NOTE: Old pre-v1 without a system prompt)",
/* .template_str= */ "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}",
/* .expected_output= */ "[INST] You are a helpful assistant\nHello [/INST]Hi there</s>[INST] Who are you [/INST] I am an assistant </s>[INST] Another question [/INST]",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "<s>",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "TheBloke/FusionNet_34Bx2_MoE-AWQ",
/* .template_str= */ "{%- for idx in range(0, messages|length) -%}\n{%- if messages[idx]['role'] == 'user' -%}\n{%- if idx > 1 -%}\n{{- bos_token + '[INST] ' + messages[idx]['content'] + ' [/INST]' -}}\n{%- else -%}\n{{- messages[idx]['content'] + ' [/INST]' -}}\n{%- endif -%}\n{% elif messages[idx]['role'] == 'system' %}\n{{- '[INST] <<SYS>>\\n' + messages[idx]['content'] + '\\n<</SYS>>\\n\\n' -}}\n{%- elif messages[idx]['role'] == 'assistant' -%}\n{{- ' ' + messages[idx]['content'] + ' ' + eos_token -}}\n{% endif %}\n{% endfor %}",
/* .expected_output= */ "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST]Hi there</s><s>[INST] Who are you [/INST] I am an assistant </s><s>[INST] Another question [/INST]",
/* .expected_output_jinja= */ "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST] Hi there </s><s>[INST] Who are you [/INST] I am an assistant </s><s>[INST] Another question [/INST]",
/* .bos_token= */ "<s>",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "bofenghuang/vigogne-2-70b-chat",
/* .template_str= */ "{{ bos_token }}{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% elif true == true and not '<<SYS>>' in messages[0]['content'] %}{% set loop_messages = messages %}{% set system_message = 'Vous êtes Vigogne, un assistant IA créé par Zaion Lab. Vous suivez extrêmement bien les instructions. Aidez autant que vous le pouvez.' %}{% else %}{% set loop_messages = messages %}{% set system_message = false %}{% endif %}{% for message in loop_messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if loop.index0 == 0 and system_message != false %}{% set content = '<<SYS>>\\n' + system_message + '\\n<</SYS>>\\n\\n' + message['content'] %}{% else %}{% set content = message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + content.strip() + ' [/INST]' }}{% elif message['role'] == 'system' %}{{ '<<SYS>>\\n' + content.strip() + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content.strip() + ' ' + eos_token }}{% endif %}{% endfor %}",
/* .expected_output= */ "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST]Hi there</s>[INST] Who are you [/INST]I am an assistant</s>[INST] Another question [/INST]",
/* .expected_output_jinja= */ "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST] Hi there </s>[INST] Who are you [/INST] I am an assistant </s>[INST] Another question [/INST]",
/* .bos_token= */ "",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "mlabonne/AlphaMonarch-7B",
/* .template_str= */ "{% for message in messages %}{{bos_token + message['role'] + '\\n' + message['content'] + eos_token + '\\n'}}{% endfor %}{% if add_generation_prompt %}{{ bos_token + 'assistant\\n' }}{% endif %}",
/* .expected_output= */ "system\nYou are a helpful assistant</s>\n<s>user\nHello</s>\n<s>assistant\nHi there</s>\n<s>user\nWho are you</s>\n<s>assistant\n I am an assistant </s>\n<s>user\nAnother question</s>\n<s>assistant\n",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "<s>",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "google/gemma-7b-it",
/* .template_str= */ "{% if messages[0]['role'] == 'system' %}{{ raise_exception('System role not supported') }}{% endif %}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if (message['role'] == 'assistant') %}{% set role = 'model' %}{% else %}{% set role = message['role'] %}{% endif %}{{ '<start_of_turn>' + role + '\\n' + message['content'] | trim + '<end_of_turn>\\n' }}{% endfor %}{% if add_generation_prompt %}{{'<start_of_turn>model\\n'}}{% endif %}",
/* .expected_output= */ "<start_of_turn>user\nYou are a helpful assistant\n\nHello<end_of_turn>\n<start_of_turn>model\nHi there<end_of_turn>\n<start_of_turn>user\nWho are you<end_of_turn>\n<start_of_turn>model\nI am an assistant<end_of_turn>\n<start_of_turn>user\nAnother question<end_of_turn>\n<start_of_turn>model\n",
/* .expected_output_jinja= */ "<start_of_turn>user\nYou are a helpful assistant\nHello<end_of_turn>\n<start_of_turn>model\nHi there<end_of_turn>\n<start_of_turn>user\nWho are you<end_of_turn>\n<start_of_turn>model\nI am an assistant<end_of_turn>\n<start_of_turn>user\nAnother question<end_of_turn>\n<start_of_turn>model\n",
},
{
/* .name= */ "OrionStarAI/Orion-14B-Chat",
/* .template_str= */ "{% for message in messages %}{% if loop.first %}{{ bos_token }}{% endif %}{% if message['role'] == 'user' %}{{ 'Human: ' + message['content'] + '\\n\\nAssistant: ' + eos_token }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token }}{% endif %}{% endfor %}",
/* .expected_output= */ "Human: You are a helpful assistant\n\nHello\n\nAssistant: </s>Hi there</s>Human: Who are you\n\nAssistant: </s> I am an assistant </s>Human: Another question\n\nAssistant: </s>",
/* .expected_output_jinja= */ "Human: You are a helpful assistant\nHello\n\nAssistant: </s>Hi there</s>Human: Who are you\n\nAssistant: </s> I am an assistant </s>Human: Another question\n\nAssistant: ",
/* .bos_token= */ "",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "openchat/openchat-3.5-0106",
// The included chat_template differs from the author's suggestions here: https://huggingface.co/openchat/openchat_3.5/discussions/5#65448109b4a3f3a2f486fd9d
// So we match against the included template but implement the suggested version.
/* .template_str= */ "{{ bos_token }}{% for message in messages %}{{ 'GPT4 Correct ' + message['role'].title() + ': ' + message['content'] + '<|end_of_turn|>'}}{% endfor %}{% if add_generation_prompt %}{{ 'GPT4 Correct Assistant:' }}{% endif %}",
/* .expected_output= */ "You are a helpful assistant<|end_of_turn|>GPT4 Correct User: Hello<|end_of_turn|>GPT4 Correct Assistant: Hi there<|end_of_turn|>GPT4 Correct User: Who are you<|end_of_turn|>GPT4 Correct Assistant: I am an assistant <|end_of_turn|>GPT4 Correct User: Another question<|end_of_turn|>GPT4 Correct Assistant:",
/* .expected_output_jinja= */ "GPT4 Correct System: You are a helpful assistant<|end_of_turn|>GPT4 Correct User: Hello<|end_of_turn|>GPT4 Correct Assistant: Hi there<|end_of_turn|>GPT4 Correct User: Who are you<|end_of_turn|>GPT4 Correct Assistant: I am an assistant <|end_of_turn|>GPT4 Correct User: Another question<|end_of_turn|>GPT4 Correct Assistant:",
},
{
/* .name= */ "deepseek-ai/deepseek-coder-33b-instruct",
/* .template_str= */ "{% if not add_generation_prompt is defined %}\n{% set add_generation_prompt = false %}\n{% endif %}\n{%- set ns = namespace(found=false) -%}\n{%- for message in messages -%}\n {%- if message['role'] == 'system' -%}\n {%- set ns.found = true -%}\n {%- endif -%}\n{%- endfor -%}\n{{bos_token}}{%- if not ns.found -%}\n{{'You are an AI programming assistant, utilizing the Deepseek Coder model, developed by Deepseek Company, and you only answer questions related to computer science. For politically sensitive questions, security and privacy issues, and other non-computer science questions, you will refuse to answer\\n'}}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'system' %}\n{{ message['content'] }}\n {%- else %}\n {%- if message['role'] == 'user' %}\n{{'### Instruction:\\n' + message['content'] + '\\n'}}\n {%- else %}\n{{'### Response:\\n' + message['content'] + '\\n<|EOT|>\\n'}}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{% if add_generation_prompt %}\n{{'### Response:'}}\n{% endif %}",
/* .expected_output= */ "You are a helpful assistant### Instruction:\nHello\n### Response:\nHi there\n<|EOT|>\n### Instruction:\nWho are you\n### Response:\n I am an assistant \n<|EOT|>\n### Instruction:\nAnother question\n### Response:\n",
/* .expected_output_jinja= */ "",
},
{
/* .name= */ "eachadea/vicuna-13b-1.1",
// No template included in tokenizer_config.json, so this template likely needs to be manually set.
/* .template_str= */ "{%- for message in messages %}{%- if message['role'] == 'system' -%}{{- '' + message['content'] + '\n\n' -}}{%- else -%}{%- if message['role'] == 'user' -%}{{-'USER: ' + message['content'] + '\n'-}}{%- else -%}{{-'ASSISTANT: ' + message['content'] + '</s>\n' -}}{%- endif -%}{%- endif -%}{%- endfor -%}{%- if add_generation_prompt -%}{{-'ASSISTANT:'-}}{%- endif -%}",
/* .expected_output= */ "You are a helpful assistant\n\nUSER: Hello\nASSISTANT: Hi there</s>\nUSER: Who are you\nASSISTANT: I am an assistant </s>\nUSER: Another question\nASSISTANT:",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
{
/* .name= */ "Orca-Vicuna",
// No template included in tokenizer_config.json, so this template likely needs to be manually set.
/* .template_str= */ "{%- for message in messages %}{%- if message['role'] == 'system' -%}{{-'SYSTEM: ' + message['content'] + '\n' -}}{%- else -%}{%- if message['role'] == 'user' -%}{{-'USER: ' + message['content'] + '\n'-}}{%- else -%}{{-'ASSISTANT: ' + message['content'] + '</s>\n' -}}{%- endif -%}{%- endif -%}{%- endfor -%}{%- if add_generation_prompt -%}{{-'ASSISTANT:'-}}{%- endif -%}",
/* .expected_output= */ "SYSTEM: You are a helpful assistant\nUSER: Hello\nASSISTANT: Hi there</s>\nUSER: Who are you\nASSISTANT: I am an assistant </s>\nUSER: Another question\nASSISTANT:",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
{
/* .name= */ "CohereForAI/c4ai-command-r-plus",
/* .template_str= */ "{{ bos_token }}{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% elif false == true %}{% set loop_messages = messages %}{% set system_message = 'You are Command-R, a brilliant, sophisticated, AI-assistant trained to assist human users by providing thorough responses. You are trained by Cohere.' %}{% else %}{% set loop_messages = messages %}{% set system_message = false %}{% endif %}{% if system_message != false %}{{ '<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>' + system_message + '<|END_OF_TURN_TOKEN|>' }}{% endif %}{% for message in loop_messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<|START_OF_TURN_TOKEN|><|USER_TOKEN|>' + content.strip() + '<|END_OF_TURN_TOKEN|>' }}{% elif message['role'] == 'assistant' %}{{ '<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>' + content.strip() + '<|END_OF_TURN_TOKEN|>' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>' }}{% endif %}",
/* .expected_output= */ "<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>You are a helpful assistant<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>Hi there<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Who are you<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>I am an assistant<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Another question<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>",
/* .expected_output_jinja= */ "",
},
{
/* .name= */ "Llama-3",
/* .template_str= */ "{% set loop_messages = messages %}{% for message in loop_messages %}{% set content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' %}{% if loop.index0 == 0 %}{% set content = bos_token + content %}{% endif %}{{ content }}{% endfor %}{{ '<|start_header_id|>assistant<|end_header_id|>\n\n' }}",
/* .expected_output= */ "<|start_header_id|>system<|end_header_id|>\n\nYou are a helpful assistant<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nHello<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\nHi there<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nWho are you<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\nI am an assistant<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nAnother question<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
/* .expected_output_jinja= */ "",
},
{
/* .name= */ "Phi-3-mini",
/* .template_str= */ "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') %}{{'<|user|>' + '\n' + message['content'] + '<|end|>' + '\n' + '<|assistant|>' + '\n'}}{% elif (message['role'] == 'assistant') %}{{message['content'] + '<|end|>' + '\n'}}{% endif %}{% endfor %}",
/* .expected_output= */ "<|system|>\nYou are a helpful assistant<|end|>\n<|user|>\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
/* .expected_output_jinja= */ "<|user|>\nYou are a helpful assistant\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
},
{
/* .name= */ "Phi-3-small",
/* .template_str= */ "{{ bos_token }}{% for message in messages %}{{'<|' + message['role'] + '|>' + '\n' + message['content'] + '<|end|>\n' }}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>\n' }}{% else %}{{ eos_token }}{% endif %}",
/* .expected_output= */ "<|system|>\nYou are a helpful assistant<|end|>\n<|user|>\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
/* .expected_output_jinja= */ "",
},
{
/* .name= */ "Phi-3-medium",
/* .template_str= */ "{% for message in messages %}{% if (message['role'] == 'user') %}{{'<|user|>' + '\n' + message['content'] + '<|end|>' + '\n' + '<|assistant|>' + '\n'}}{% elif (message['role'] == 'assistant') %}{{message['content'] + '<|end|>' + '\n'}}{% endif %}{% endfor %}",
/* .expected_output= */ "<|system|>\nYou are a helpful assistant<|end|>\n<|user|>\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
/* .expected_output_jinja= */ "<|user|>\nYou are a helpful assistant\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
},
{
/* .name= */ "Phi-3-vision",
/* .template_str= */ "{% for message in messages %}{{'<|' + message['role'] + '|>' + '\n' + message['content'] + '<|end|>\n' }}{% endfor %}{% if add_generation_prompt and messages[-1]['role'] != 'assistant' %}{{- '<|assistant|>\n' -}}{% endif %}",
/* .expected_output= */ "<|system|>\nYou are a helpful assistant<|end|>\n<|user|>\nHello<|end|>\n<|assistant|>\nHi there<|end|>\n<|user|>\nWho are you<|end|>\n<|assistant|>\n I am an assistant <|end|>\n<|user|>\nAnother question<|end|>\n<|assistant|>\n",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
{
/* .name= */ "ChatGLM3",
/* .template_str= */ "{% for message in messages %}{% if loop.first %}[gMASK]sop<|{{ message['role'] }}|>\n {{ message['content'] }}{% else %}<|{{ message['role'] }}|>\n {{ message['content'] }}{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>{% endif %}",
/* .expected_output= */ "[gMASK]sop<|system|>\n You are a helpful assistant<|user|>\n Hello<|assistant|>\n Hi there<|user|>\n Who are you<|assistant|>\n I am an assistant <|user|>\n Another question<|assistant|>",
/* .expected_output_jinja= */ "[gMASK]sop<|system|>\nYou are a helpful assistant<|user|>\nHello<|assistant|>\nHi there<|user|>\nWho are you<|assistant|>\n I am an assistant <|user|>\nAnother question<|assistant|>",
},
{
/* .name= */ "ChatGLM4",
/* .template_str= */ u8"[gMASK]<sop>{% for item in messages %}{% if item['tools'] is defined %}<|system|>\n你是一个名为 ChatGLM 的人工智能助手。你是基于智谱AI训练的语言模型 GLM-4 模型开发的，你的任务是针对用户的问题和要求提供适当的答复和支持。\n\n# 可用工具{% set tools = item['tools'] %}{% for tool in tools %}{% if tool['type'] == 'function' %}\n\n## {{ tool['function']['name'] }}\n\n{{ tool['function'] | tojson(indent=4) }}\n......{% endif %}{% endfor %}{% endif %}{% if item['content'] %}<|{{ item['role'] }}|>{{ item['metadata'] }}\n{{ item['content'] }}{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>{% endif %}",
/* .expected_output= */ "[gMASK]<sop><|system|>\nYou are a helpful assistant<|user|>\nHello<|assistant|>\nHi there<|user|>\nWho are you<|assistant|>\n I am an assistant <|user|>\nAnother question<|assistant|>",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
{
/* .name= */ "GLMEdge",
/* .template_str= */ "{% for item in messages %}{% if item['role'] == 'system' %}<|system|>\n{{ item['content'] }}{% elif item['role'] == 'user' %}<|user|>\n{{ item['content'] }}{% elif item['role'] == 'assistant' %}<|assistant|>\n{{ item['content'] }}{% endif %}{% endfor %}<|assistant|>",
/* .expected_output= */ "<|system|>\nYou are a helpful assistant<|user|>\nHello<|assistant|>\nHi there<|user|>\nWho are you<|assistant|>\n I am an assistant <|user|>\nAnother question<|assistant|>",
/* .expected_output_jinja= */ "<|system|>\nYou are a helpful assistant<|user|>\nHello<|assistant|>\nHi there<|user|>\nWho are you<|assistant|>\n I am an assistant <|user|>\nAnother question<|assistant|>",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
{
/* .name= */ "MiniCPM-3B-OpenHermes-2.5-v2-GGUF",
/* .template_str= */ u8"{% for message in messages %}{% if message['role'] == 'user' %}{{'<用户>' + message['content'].strip() + '<AI>'}}{% else %}{{message['content'].strip()}}{% endif %}{% endfor %}",
/* .expected_output= */ u8"You are a helpful assistant<用户>Hello<AI>Hi there<用户>Who are you<AI>I am an assistant<用户>Another question<AI>",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
{
/* .name= */ "DeepSeek-V2",
/* .template_str= */ "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ 'User: ' + message['content'] + '\n\n' }}{% elif message['role'] == 'assistant' %}{{ 'Assistant: ' + message['content'] + eos_token }}{% elif message['role'] == 'system' %}{{ message['content'] + '\n\n' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ 'Assistant:' }}{% endif %}",
/* .expected_output= */ u8"You are a helpful assistant\n\nUser: Hello\n\nAssistant: Hi there<｜end▁of▁sentence｜>User: Who are you\n\nAssistant: I am an assistant <｜end▁of▁sentence｜>User: Another question\n\nAssistant:",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "<｜end▁of▁sentence｜>",
},
{
/* .name= */ "ibm-granite/granite-3.0-8b-instruct",
/* .template_str= */ "{%- if tools %}\n {{- '<|start_of_role|>available_tools<|end_of_role|>\n' }}\n {%- for tool in tools %}\n {{- tool | tojson(indent=4) }}\n {%- if not loop.last %}\n {{- '\n\n' }}\n {%- endif %}\n {%- endfor %}\n {{- '<|end_of_text|>\n' }}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'system' %}\n {{- '<|start_of_role|>system<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'user' %}\n {{- '<|start_of_role|>user<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'assistant' %}\n {{- '<|start_of_role|>assistant<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'assistant_tool_call' %}\n {{- '<|start_of_role|>assistant<|end_of_role|><|tool_call|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'tool_response' %}\n {{- '<|start_of_role|>tool_response<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- endif %}\n {%- if loop.last and add_generation_prompt %}\n {{- '<|start_of_role|>assistant<|end_of_role|>' }}\n {%- endif %}\n{%- endfor %}",
/* .expected_output= */ "<|start_of_role|>system<|end_of_role|>You are a helpful assistant<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Hello<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>Hi there<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Who are you<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|> I am an assistant <|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Another question<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>\n",
/* .expected_output_jinja= */ "<|start_of_role|>system<|end_of_role|>You are a helpful assistant<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Hello<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>Hi there<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Who are you<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|> I am an assistant <|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Another question<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>",
},
{
/* .name= */ "mistralai/Mistral-7B-Instruct-v0.2 (mistralai 'v1' template with a system prompt)",
/* .template_str= */ "{%- if messages[0]['role'] == 'system' %}\n {%- set system_message = messages[0]['content'] %}\n {%- set loop_messages = messages[1:] %}\n{%- else %}\n {%- set loop_messages = messages %}\n{%- endif %}\n\n{{- bos_token }}\n{%- for message in loop_messages %}\n {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}\n {{- raise_exception('After the optional system message, conversation roles must alternate user/assistant/user/assistant/...') }}\n {%- endif %}\n {%- if message['role'] == 'user' %}\n {%- if loop.first and system_message is defined %}\n {{- ' [INST] ' + system_message + '\\n\\n' + message['content'] + ' [/INST]' }}\n {%- else %}\n {{- ' [INST] ' + message['content'] + ' [/INST]' }}\n {%- endif %}\n {%- elif message['role'] == 'assistant' %}\n {{- ' ' + message['content'] + eos_token}}\n {%- else %}\n {{- raise_exception('Only user and assistant roles are supported, with the exception of an initial optional system message!') }}\n {%- endif %}\n{%- endfor %}\n",
/* .expected_output= */ " [INST] You are a helpful assistant\n\nHello [/INST] Hi there</s> [INST] Who are you [/INST] I am an assistant </s> [INST] Another question [/INST]",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "Mistral-Large-Instruct-2407 (mistralai 'v3' template; modified to have system prompt at start)",
/* .template_str= */ "{%- if messages[0][\"role\"] == \"system\" %}\n {%- set system_message = messages[0][\"content\"] %}\n {%- set loop_messages = messages[1:] %}\n{%- else %}\n {%- set loop_messages = messages %}\n{%- endif %}\n{%- if not tools is defined %}\n {%- set tools = none %}\n{%- endif %}\n{%- set user_messages = loop_messages | selectattr(\"role\", \"equalto\", \"user\") | list %}\n\n{#- This block checks for alternating user/assistant messages, skipping tool calling messages #}\n{%- set ns = namespace() %}\n{%- set ns.index = 0 %}\n{%- for message in loop_messages %}\n {%- if not (message.role == \"tool\" or message.role == \"tool_results\" or (message.tool_calls is defined and message.tool_calls is not none)) %}\n {%- if (message[\"role\"] == \"user\") != (ns.index % 2 == 0) %}\n {{- raise_exception(\"After the optional system message, conversation roles must alternate user/assistant/user/assistant/...\") }}\n {%- endif %}\n {%- set ns.index = ns.index + 1 %}\n {%- endif %}\n{%- endfor %}\n\n{{- bos_token }}\n{%- for message in loop_messages %}\n {%- if message[\"role\"] == \"user\" %}\n {%- if tools is not none and (message == user_messages[-1]) %}\n {{- \"[AVAILABLE_TOOLS] [\" }}\n {%- for tool in tools %}\n {%- set tool = tool.function %}\n {{- '{\"type\": \"function\", \"function\": {' }}\n {%- for key, val in tool.items() if key != \"return\" %}\n {%- if val is string %}\n {{- '\"' + key + '\": \"' + val + '\"' }}\n {%- else %}\n {{- '\"' + key + '\": ' + val|tojson }}\n {%- endif %}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- endif %}\n {%- endfor %}\n {{- \"}}\" }}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- else %}\n {{- \"]\" }}\n {%- endif %}\n {%- endfor %}\n {{- \"[/AVAILABLE_TOOLS]\" }}\n {%- endif %}\n {%- if loop.last and system_message is defined %}\n {{- \"[INST] \" + system_message + \"\\n\\n\" + message[\"content\"] + \"[/INST]\" }}\n {%- else %}\n {{- \"[INST] \" + message[\"content\"] + \"[/INST]\" }}\n {%- endif %}\n {%- elif message.tool_calls is defined and message.tool_calls is not none %}\n {{- \"[TOOL_CALLS] [\" }}\n {%- for tool_call in message.tool_calls %}\n {%- set out = tool_call.function|tojson %}\n {{- out[:-1] }}\n {%- if not tool_call.id is defined or tool_call.id|length != 9 %}\n {{- raise_exception(\"Tool call IDs should be alphanumeric strings with length 9!\") }}\n {%- endif %}\n {{- ', \"id\": \"' + tool_call.id + '\"}' }}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- else %}\n {{- \"]\" + eos_token }}\n {%- endif %}\n {%- endfor %}\n {%- elif message[\"role\"] == \"assistant\" %}\n {{- \" \" + message[\"content\"]|trim + eos_token}}\n {%- elif message[\"role\"] == \"tool_results\" or message[\"role\"] == \"tool\" %}\n {%- if message.content is defined and message.content.content is defined %}\n {%- set content = message.content.content %}\n {%- else %}\n {%- set content = message.content %}\n {%- endif %}\n {{- '[TOOL_RESULTS] {\"content\": ' + content|string + \", \" }}\n {%- if not message.tool_call_id is defined or message.tool_call_id|length != 9 %}\n {{- raise_exception(\"Tool call IDs should be alphanumeric strings with length 9!\") }}\n {%- endif %}\n {{- '\"call_id\": \"' + message.tool_call_id + '\"}[/TOOL_RESULTS]' }}\n {%- else %}\n {{- raise_exception(\"Only user and assistant roles are supported, with the exception of an initial optional system message!\") }}\n {%- endif %}\n{%- endfor %}\n",
/* .expected_output= */ "[INST] You are a helpful assistant\n\nHello[/INST] Hi there</s>[INST] Who are you[/INST] I am an assistant</s>[INST] Another question[/INST]",
/* .expected_output_jinja= */ "[INST] Hello[/INST] Hi there</s>[INST] Who are you[/INST] I am an assistant</s>[INST] You are a helpful assistant\n\nAnother question[/INST]",
/* .bos_token= */ "",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "Mistral-Nemo-Instruct-2407 (mistralai 'v3-tekken' template; modified to have system prompt at start)",
/* .template_str= */ "{%- if messages[0][\"role\"] == \"system\" %}\n {%- set system_message = messages[0][\"content\"] %}\n {%- set loop_messages = messages[1:] %}\n{%- else %}\n {%- set loop_messages = messages %}\n{%- endif %}\n{%- if not tools is defined %}\n {%- set tools = none %}\n{%- endif %}\n{%- set user_messages = loop_messages | selectattr(\"role\", \"equalto\", \"user\") | list %}\n\n{#- This block checks for alternating user/assistant messages, skipping tool calling messages #}\n{%- set ns = namespace() %}\n{%- set ns.index = 0 %}\n{%- for message in loop_messages %}\n {%- if not (message.role == \"tool\" or message.role == \"tool_results\" or (message.tool_calls is defined and message.tool_calls is not none)) %}\n {%- if (message[\"role\"] == \"user\") != (ns.index % 2 == 0) %}\n {{- raise_exception(\"After the optional system message, conversation roles must alternate user/assistant/user/assistant/...\") }}\n {%- endif %}\n {%- set ns.index = ns.index + 1 %}\n {%- endif %}\n{%- endfor %}\n\n{{- bos_token }}\n{%- for message in loop_messages %}\n {%- if message[\"role\"] == \"user\" %}\n {%- if tools is not none and (message == user_messages[-1]) %}\n {{- \"[AVAILABLE_TOOLS][\" }}\n {%- for tool in tools %}\n {%- set tool = tool.function %}\n {{- '{\"type\": \"function\", \"function\": {' }}\n {%- for key, val in tool.items() if key != \"return\" %}\n {%- if val is string %}\n {{- '\"' + key + '\": \"' + val + '\"' }}\n {%- else %}\n {{- '\"' + key + '\": ' + val|tojson }}\n {%- endif %}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- endif %}\n {%- endfor %}\n {{- \"}}\" }}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- else %}\n {{- \"]\" }}\n {%- endif %}\n {%- endfor %}\n {{- \"[/AVAILABLE_TOOLS]\" }}\n {%- endif %}\n {%- if loop.last and system_message is defined %}\n {{- \"[INST]\" + system_message + \"\\n\\n\" + message[\"content\"] + \"[/INST]\" }}\n {%- else %}\n {{- \"[INST]\" + message[\"content\"] + \"[/INST]\" }}\n {%- endif %}\n {%- elif (message.tool_calls is defined and message.tool_calls is not none) %}\n {{- \"[TOOL_CALLS][\" }}\n {%- for tool_call in message.tool_calls %}\n {%- set out = tool_call.function|tojson %}\n {{- out[:-1] }}\n {%- if not tool_call.id is defined or tool_call.id|length != 9 %}\n {{- raise_exception(\"Tool call IDs should be alphanumeric strings with length 9!\") }}\n {%- endif %}\n {{- ', \"id\": \"' + tool_call.id + '\"}' }}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- else %}\n {{- \"]\" + eos_token }}\n {%- endif %}\n {%- endfor %}\n {%- elif message[\"role\"] == \"assistant\" %}\n {{- message[\"content\"] + eos_token}}\n {%- elif message[\"role\"] == \"tool_results\" or message[\"role\"] == \"tool\" %}\n {%- if message.content is defined and message.content.content is defined %}\n {%- set content = message.content.content %}\n {%- else %}\n {%- set content = message.content %}\n {%- endif %}\n {{- '[TOOL_RESULTS]{\"content\": ' + content|string + \", \" }}\n {%- if not message.tool_call_id is defined or message.tool_call_id|length != 9 %}\n {{- raise_exception(\"Tool call IDs should be alphanumeric strings with length 9!\") }}\n {%- endif %}\n {{- '\"call_id\": \"' + message.tool_call_id + '\"}[/TOOL_RESULTS]' }}\n {%- else %}\n {{- raise_exception(\"Only user and assistant roles are supported, with the exception of an initial optional system message!\") }}\n {%- endif %}\n{%- endfor %}\n",
/* .expected_output= */ "[INST]You are a helpful assistant\n\nHello[/INST]Hi there</s>[INST]Who are you[/INST] I am an assistant </s>[INST]Another question[/INST]",
/* .expected_output_jinja= */ "[INST]Hello[/INST]Hi there</s>[INST]Who are you[/INST] I am an assistant </s>[INST]You are a helpful assistant\n\nAnother question[/INST]",
/* .bos_token= */ "",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "mistralai/Mistral-Large-Instruct-2411 (mistralai 'v7' template)",
/* .template_str= */ "{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + '[/INST]' }}{% elif message['role'] == 'system' %}{{ '[SYSTEM_PROMPT] ' + message['content'] + '[/SYSTEM_PROMPT]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + eos_token }}{% else %}{{ raise_exception('Only user, system and assistant roles are supported!') }}{% endif %}{% endfor %}",
/* .expected_output= */ "[SYSTEM_PROMPT] You are a helpful assistant[/SYSTEM_PROMPT][INST] Hello[/INST] Hi there</s>[INST] Who are you[/INST] I am an assistant </s>[INST] Another question[/INST]",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "</s>",
},
{
/* .name= */ "ai-sage/GigaChat-20B-A3B-instruct",
/* .template_str= */ "{% if messages[0]['role'] == 'system' -%}\n {%- set loop_messages = messages[1:] -%}\n {%- set system_message = bos_token + messages[0]['content'] + additional_special_tokens[1] -%}\n{%- else -%}\n {%- set loop_messages = messages -%}\n {%- set system_message = bos_token + '' -%}\n{%- endif -%}\n{%- for message in loop_messages %}\n {% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}\n {{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}\n {% endif %}\n \n {%- if loop.index0 == 0 -%}\n {{ system_message -}}\n {%- endif -%}\n {%- if message['role'] == 'user' -%}\n {{ message['role'] + additional_special_tokens[0] + message['content'] + additional_special_tokens[1] -}}\n {{ 'available functions' + additional_special_tokens[0] + additional_special_tokens[2] + additional_special_tokens[3] + additional_special_tokens[1] -}}\n {%- endif -%}\n {%- if message['role'] == 'assistant' -%}\n {{ message['role'] + additional_special_tokens[0] + message['content'] + additional_special_tokens[1] -}}\n {%- endif -%}\n {%- if loop.last and add_generation_prompt -%}\n {{ 'assistant' + additional_special_tokens[0] -}}\n {%- endif -%}\n{%- endfor %}",
/* .expected_output= */ "<s>You are a helpful assistant<|message_sep|>user<|role_sep|>Hello<|message_sep|>available functions<|role_sep|>[]<|message_sep|>assistant<|role_sep|>Hi there<|message_sep|>user<|role_sep|>Who are you<|message_sep|>available functions<|role_sep|>[]<|message_sep|>assistant<|role_sep|> I am an assistant <|message_sep|>user<|role_sep|>Another question<|message_sep|>available functions<|role_sep|>[]<|message_sep|>assistant<|role_sep|>",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
/* .supported_with_jinja= */ false, // Requires additional_special_tokens as extra context
},
{
/* .name= */ "Infinigence/Megrez-3B-Instruct",
/* .template_str= */ u8"{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|role_start|>system<|role_end|>你是Megrez-3B-Instruct，将针对用户的问题给出详细的、积极的回答。<|turn_end|>' }}{% endif %}{{ '<|role_start|>' + message['role'] + '<|role_end|>' + message['content'] + '<|turn_end|>' }}{% endfor %}{% if add_generation_prompt %}{{ '<|role_start|>assistant<|role_end|>' }}{% endif %}",
/* .expected_output= */ "<|role_start|>system<|role_end|>You are a helpful assistant<|turn_end|><|role_start|>user<|role_end|>Hello<|turn_end|><|role_start|>assistant<|role_end|>Hi there<|turn_end|><|role_start|>user<|role_end|>Who are you<|turn_end|><|role_start|>assistant<|role_end|> I am an assistant <|turn_end|><|role_start|>user<|role_end|>Another question<|turn_end|><|role_start|>assistant<|role_end|>",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
{
/* .name= */ "phi-4",
/* .template_str= */ "{% for message in messages %}{% if (message['role'] == 'system') %}{{'<|im_start|>system<|im_sep|>' + message['content'] + '<|im_end|>'}}{% elif (message['role'] == 'user') %}{{'<|im_start|>user<|im_sep|>' + message['content'] + '<|im_end|><|im_start|>assistant<|im_sep|>'}}{% elif (message['role'] == 'assistant') %}{{message['content'] + '<|im_end|>'}}{% endif %}{% endfor %}",
/* .expected_output= */ "<|im_start|>system<|im_sep|>You are a helpful assistant<|im_end|><|im_start|>user<|im_sep|>Hello<|im_end|><|im_start|>assistant<|im_sep|>Hi there<|im_end|><|im_start|>user<|im_sep|>Who are you<|im_end|><|im_start|>assistant<|im_sep|> I am an assistant <|im_end|><|im_start|>user<|im_sep|>Another question<|im_end|><|im_start|>assistant<|im_sep|>",
/* .expected_output_jinja= */ "",
/* .bos_token= */ "",
/* .eos_token= */ "",
},
};
std::vector<char> formatted_chat(1024);
int32_t res;
// test invalid chat template
res = llama_chat_apply_template(nullptr, "INVALID TEMPLATE", conversation, message_count, true, formatted_chat.data(), formatted_chat.size());
assert(res < 0);
// list all supported templates
std::vector<const char *> supported_tmpl;
res = llama_chat_builtin_templates(nullptr, 0);
assert(res > 0);
supported_tmpl.resize(res);
res = llama_chat_builtin_templates(supported_tmpl.data(), supported_tmpl.size());
printf("Built-in chat templates:\n");
for (auto tmpl : supported_tmpl) {
printf(" %s\n", tmpl);
}
for (size_t i = 0; i < templates.size(); i++) {
std::string custom_template = templates[i];
std::string expected = expected_output[i];
// test invalid chat template
res = llama_chat_apply_template("INVALID TEMPLATE", conversation.data(), conversation.size(), true, formatted_chat.data(), formatted_chat.size());
assert(res < 0);
const auto add_generation_prompt = true;
for (const auto & test_case : test_cases) {
printf("\n\n=== %s ===\n\n", test_case.name.c_str());
formatted_chat.resize(1024);
res = llama_chat_apply_template(
nullptr,
custom_template.c_str(),
conversation,
message_count,
true,
test_case.template_str.c_str(),
conversation.data(),
conversation.size(),
add_generation_prompt,
formatted_chat.data(),
formatted_chat.size()
);
formatted_chat.resize(res);
std::string output(formatted_chat.data(), formatted_chat.size());
printf("%s\n", output.c_str());
printf("-------------------------\n");
assert(output == expected);
if (output != test_case.expected_output) {
printf("Expected:\n%s\n", test_case.expected_output.c_str());
printf("-------------------------\n");
printf("Actual:\n%s\n", output.c_str());
fflush(stdout);
assert(output == test_case.expected_output);
}
}
std::vector<common_chat_msg> messages;
for (const auto & msg : conversation) {
messages.push_back(simple_msg(msg.role, msg.content));
}
for (const auto & test_case : test_cases) {
if (!test_case.supported_with_jinja) {
continue;
}
printf("\n\n=== %s (jinja) ===\n\n", test_case.name.c_str());
try {
auto tmpls = common_chat_templates_init(/* model= */ nullptr, test_case.template_str.c_str(), test_case.bos_token, test_case.eos_token);
common_chat_templates_inputs inputs;
inputs.use_jinja = true;
inputs.messages = messages;
inputs.add_generation_prompt = add_generation_prompt;
auto output = common_chat_templates_apply(tmpls.get(), inputs).prompt;
output = normalize_newlines(output);
auto expected_output = normalize_newlines(test_case.expected_output_jinja.empty() ? test_case.expected_output : test_case.expected_output_jinja);
if (output != expected_output) {
printf("Expected:\n%s\n", expected_output.c_str());
printf("-------------------------\n");
printf("Actual:\n%s\n", output.c_str());
fflush(stdout);
assert(output == expected_output);
}
} catch (const std::exception & e) {
printf("ERROR: %s\n", e.what());
assert(false);
}
}
// test llama_chat_format_single for system message
printf("\n\n=== llama_chat_format_single (system message) ===\n\n");
std::vector<common_chat_msg> chat2;
common_chat_msg sys_msg{"system", "You are a helpful assistant"};
auto sys_msg = simple_msg("system", "You are a helpful assistant");
auto fmt_sys = [&](std::string tmpl) {
auto output = common_chat_format_single(nullptr, tmpl, chat2, sys_msg, false);
printf("fmt_sys(%s) : %s\n", tmpl.c_str(), output.c_str());
auto fmt_sys = [&](std::string tmpl_str) {
auto tmpls = common_chat_templates_init(/* model= */ nullptr, tmpl_str);
auto output = common_chat_format_single(tmpls.get(), chat2, sys_msg, false, /* use_jinja= */ false);
printf("fmt_sys(%s) : %s\n", tmpl_str.c_str(), output.c_str());
printf("-------------------------\n");
return output;
};
assert(fmt_sys("chatml") == "<|im_start|>system\nYou are a helpful assistant<|im_end|>\n");
assert(fmt_sys("mistral-v1") == " [INST] You are a helpful assistant\n\n");
assert(fmt_sys("mistral-v3") == "[INST] You are a helpful assistant\n\n");
assert(fmt_sys("mistral-v3-tekken") == "[INST]You are a helpful assistant\n\n");
assert(fmt_sys("mistral-v7") == "[SYSTEM_PROMPT] You are a helpful assistant[/SYSTEM_PROMPT]");
assert(fmt_sys("llama2") == "[INST] You are a helpful assistant\n");
assert(fmt_sys("llama2-sys") == "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\n");
assert(fmt_sys("mistral") == "[INST] You are a helpful assistant\n"); // for old pre-v1 templates
assert(fmt_sys("gemma") == ""); // for gemma, system message is merged with user message
assert(fmt_sys("llama3") == "<|start_header_id|>system<|end_header_id|>\n\nYou are a helpful assistant<|eot_id|>");
assert(fmt_sys("gigachat") == "<s>You are a helpful assistant<|message_sep|>");
// test llama_chat_format_single for user message
printf("\n\n=== llama_chat_format_single (user message) ===\n\n");
chat2.push_back({"system", "You are a helpful assistant"});
chat2.push_back({"user", "Hello"});
chat2.push_back({"assistant", "I am assistant"});
common_chat_msg new_msg{"user", "How are you"};
chat2.push_back(simple_msg("system", "You are a helpful assistant"));
chat2.push_back(simple_msg("user", "Hello"));
chat2.push_back(simple_msg("assistant", "I am assistant"));
auto new_msg = simple_msg("user", "How are you");
auto fmt_single = [&](std::string tmpl) {
auto output = common_chat_format_single(nullptr, tmpl, chat2, new_msg, true);
printf("fmt_single(%s) : %s\n", tmpl.c_str(), output.c_str());
auto fmt_single = [&](const std::string & tmpl_str) {
auto tmpls = common_chat_templates_init(/* model= */ nullptr, tmpl_str.c_str());
auto output = common_chat_format_single(tmpls.get(), chat2, new_msg, true, /* use_jinja= */ false);
printf("fmt_single(%s) : %s\n", tmpl_str.c_str(), output.c_str());
printf("-------------------------\n");
return output;
};
assert(fmt_single("chatml") == "\n<|im_start|>user\nHow are you<|im_end|>\n<|im_start|>assistant\n");
assert(fmt_single("mistral-v1") == " [INST] How are you [/INST]");
assert(fmt_single("mistral-v3") == "[INST] How are you[/INST]");
assert(fmt_single("mistral-v3-tekken") == "[INST]How are you[/INST]");
assert(fmt_single("mistral-v7") == "[INST] How are you[/INST]");
assert(fmt_single("llama2") == "[INST] How are you [/INST]");
assert(fmt_single("mistral") == "[INST] How are you [/INST]"); // for old pre-v1 templates
assert(fmt_single("gemma") == "\n<start_of_turn>user\nHow are you<end_of_turn>\n<start_of_turn>model\n");
assert(fmt_single("llama3") == "<|start_header_id|>user<|end_header_id|>\n\nHow are you<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n");
assert(fmt_single("gigachat") == "user<|role_sep|>How are you<|message_sep|>available functions<|role_sep|>[]<|message_sep|>assistant<|role_sep|>");
return 0;
}

834
tests/test-chat.cpp Normal file
View File

@@ -0,0 +1,834 @@
// Tests chat handling, including grammar generation and parsing for tool calling, for various templates.
//
// Also acts as a CLI to generate a Markdown summary of the formats of Jinja templates,
// e.g. given Minja (http://github.com/google/minja) checked out in parent dir:
//
// cmake -B build && cmake --build build --parallel && ./build/bin/test-chat ../minja/build/tests/*.jinja 2>/dev/null
//
#include <fstream>
#include <iostream>
#include <json.hpp>
#include <string>
#include "chat.h"
#include "llama-grammar.h"
#include "unicode.h"
using json = nlohmann::ordered_json;
template <class T> static void assert_equals(const T & expected, const T & actual) {
if (expected != actual) {
std::cerr << "Expected: " << expected << std::endl;
std::cerr << "Actual: " << actual << std::endl;
std::cerr << std::flush;
throw std::runtime_error("Test failed");
}
}
static std::string read_file(const std::string & path) {
std::cerr << "# Reading: " << path << '\n' << std::flush;
std::ifstream fs(path, std::ios_base::binary);
if (!fs.is_open()) {
fs = std::ifstream("../" + path, std::ios_base::binary);
if (!fs.is_open()) {
throw std::runtime_error("Failed to open file: " + path);
}
}
fs.seekg(0, std::ios_base::end);
auto size = fs.tellg();
fs.seekg(0);
std::string out;
out.resize(static_cast<size_t>(size));
fs.read(out.data(), static_cast<std::streamsize>(size));
return out;
}
static common_chat_templates_ptr read_templates(const std::string & path) {
return common_chat_templates_ptr(common_chat_templates_init(/* model= */ nullptr, read_file(path)));
}
static std::unique_ptr<llama_grammar> build_grammar(const std::string & grammar_str) {
return std::unique_ptr<llama_grammar>(
llama_grammar_init_impl(nullptr, grammar_str.c_str(), "root", false, nullptr, 0, nullptr, 0));
}
// TODO: extract to common helper (copied from test-grammar-integration.cpp)
static bool match_string(const std::string & input, llama_grammar * grammar) {
const auto cpts = unicode_cpts_from_utf8(input);
auto & stacks_cur = llama_grammar_get_stacks(grammar);
for (const auto & cpt : cpts) {
llama_grammar_accept(grammar, cpt);
if (stacks_cur.empty()) {
// no stacks means that the grammar failed to match at this point
return false;
}
}
if (std::any_of(stacks_cur.begin(), stacks_cur.end(), [](const auto & stack) { return stack.empty(); })) {
// An empty stack means that the grammar has been completed
return true;
}
return false;
}
static void assert_msg_equals(const common_chat_msg & expected, const common_chat_msg & actual) {
assert_equals(expected.role, actual.role);
assert_equals(expected.content, actual.content);
assert_equals(expected.content_parts.size(), actual.content_parts.size());
for (size_t i = 0; i < expected.content_parts.size(); i++) {
const auto & expected_part = expected.content_parts[i];
const auto & actual_part = actual.content_parts[i];
assert_equals(expected_part.type, actual_part.type);
assert_equals(expected_part.text, actual_part.text);
}
assert_equals(expected.reasoning_content, actual.reasoning_content);
assert_equals(expected.tool_calls.size(), actual.tool_calls.size());
for (size_t i = 0; i < expected.tool_calls.size(); i++) {
const auto & expected_tool_call = expected.tool_calls[i];
const auto & actual_tool_call = actual.tool_calls[i];
assert_equals(expected_tool_call.name, actual_tool_call.name);
assert_equals(json::parse(expected_tool_call.arguments).dump(), json::parse(actual_tool_call.arguments).dump());
assert_equals(expected_tool_call.id, actual_tool_call.id);
}
}
common_chat_tool special_function_tool {
/* .name = */ "special_function",
/* .description = */ "I'm special",
/* .parameters = */ R"({
"type": "object",
"properties": {
"arg1": {
"type": "integer",
"description": "The arg."
}
},
"required": ["arg1"]
})",
};
common_chat_tool python_tool {
/* .name = */ "python",
/* .description = */ "an ipython interpreter",
/* .parameters = */ R"({
"type": "object",
"properties": {
"code": {
"type": "string",
"description": "Python code to execute."
}
},
"required": ["code"]
})",
};
common_chat_tool code_interpreter_tool {
/* .name = */ "code_interpreter",
/* .description = */ "an ipython interpreter",
/* .parameters = */ R"({
"type": "object",
"properties": {
"code": {
"type": "string",
"description": "Python code to execute."
}
},
"required": ["code"]
})",
};
std::vector<common_chat_tool> tools { special_function_tool, python_tool };
std::vector<common_chat_tool> llama_3_1_tools { special_function_tool, code_interpreter_tool };
struct delta_data {
std::string delta;
common_chat_params params;
};
static delta_data init_delta(const struct common_chat_templates * tmpls, const std::vector<std::string> & end_tokens,
const common_chat_msg & user_message,
const common_chat_msg & delta_message,
const std::vector<common_chat_tool> & tools,
const common_chat_tool_choice & tool_choice,
bool think = false) {
common_chat_templates_inputs inputs;
inputs.parallel_tool_calls = true;
inputs.messages.push_back(user_message);
inputs.tools = tools;
inputs.tool_choice = tool_choice;
inputs.extract_reasoning = think;
auto params_prefix = common_chat_templates_apply(tmpls, inputs);
inputs.messages.push_back(delta_message);
inputs.add_generation_prompt = false;
auto params_full = common_chat_templates_apply(tmpls, inputs);
std::string prefix = params_prefix.prompt;
std::string full = params_full.prompt;
if (full == prefix) {
throw std::runtime_error("Full message is the same as the prefix");
}
size_t common_prefix_length = 0;
for (size_t i = 0; i < prefix.size() && i < full.size(); ++i) {
if (prefix[i] != full[i]) {
break;
}
if (prefix[i] == '<') {
// DeepSeek R1's template (as of 20250209) adds a trailing <think> if add_generation_prompt,
// but it removes thinking tags for past messages.
// The prefix and full strings diverge at <think> vs. <｜tool▁calls▁begin｜>, we avoid consuming the leading <.
continue;
}
common_prefix_length = i + 1;
}
auto delta = full.substr(common_prefix_length);
// Strip end tokens
for (const auto & end_token : end_tokens) {
// rfind to find the last occurrence
auto pos = delta.rfind(end_token);
if (pos != std::string::npos) {
delta = delta.substr(0, pos);
break;
}
}
return { delta, params_full };
}
/*
Applies the template to 1 user message w/ add_generation_prompt=true, then w/ the test message w/ add_generation_prompt=false,
gets the diff, removes any end tokens and parses the result w/ the grammar, checking that
the parsed message is the same as the test_message
*/
static void test_templates(const struct common_chat_templates * tmpls, const std::vector<std::string> & end_tokens,
const common_chat_msg & test_message,
const std::vector<common_chat_tool> & tools = {},
const std::string & expected_delta = "",
bool expect_grammar_triggered = true,
bool test_grammar_if_triggered = true,
bool think = false) {
common_chat_msg user_message;
user_message.role = "user";
user_message.content = "Hello, world!";
for (const auto & tool_choice : std::vector<common_chat_tool_choice> {COMMON_CHAT_TOOL_CHOICE_AUTO, COMMON_CHAT_TOOL_CHOICE_REQUIRED}) {
auto data = init_delta(tmpls, end_tokens, user_message, test_message, tools, tool_choice, think);
if (!expected_delta.empty()) {
assert_equals(expected_delta, data.delta);
}
if (expect_grammar_triggered) {
const auto msg = common_chat_parse(data.delta, data.params.format);
assert_msg_equals(test_message, msg);
}
if (!test_message.tool_calls.empty()) {
GGML_ASSERT(!data.params.grammar.empty());
}
if (!data.params.grammar.empty()) {
auto grammar = build_grammar(data.params.grammar);
if (!grammar) {
throw std::runtime_error("Failed to build grammar");
}
auto earliest_trigger_pos = std::string::npos;
auto constrained = data.delta;
for (const auto & trigger : data.params.grammar_triggers) {
auto pos = constrained.find(trigger.word);
if (pos == std::string::npos) {
continue;
}
if (pos > 0 && trigger.at_start) {
fprintf(stderr, "Trigger %s not at start of message, skipping:\n\n%s\n\n", trigger.word.c_str(), constrained.c_str());
continue;
}
if (earliest_trigger_pos == std::string::npos || pos < earliest_trigger_pos) {
earliest_trigger_pos = pos;
}
}
auto grammar_triggered = false;
if (earliest_trigger_pos != std::string::npos) {
constrained = constrained.substr(earliest_trigger_pos);
grammar_triggered = true;
}
if (data.params.grammar_lazy) {
assert_equals(expect_grammar_triggered, grammar_triggered);
}
if (grammar_triggered && test_grammar_if_triggered && !match_string(constrained, grammar.get())) {
throw std::runtime_error("Failed to match delta against grammar:\n\n" + data.delta +
"\n\nGrammar: " + data.params.grammar);
}
}
}
}
const common_chat_msg message_user {
"user",
"Hey there!",
/* .content_parts = */ {},
/* .tool_calls = */ {},
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_user_parts {
"user",
/* .content = */ "",
/* .content_parts = */ {
{ "text", "Hey" },
{ "text", "there" },
},
/* .tool_calls = */ {},
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist {
"assistant",
"Hello, world!\nWhat's up?",
/* .content_parts = */ {},
/* .tool_calls = */ {},
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_thoughts_unparsed_think {
"assistant",
"<think>I'm thinking</think>Hello, world!\nWhat's up?",
/* .content_parts = */ {},
/* .tool_calls = */ {},
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_thoughts_unparsed_r7b {
"assistant",
"<|START_THINKING|>I'm thinking<|END_THINKING|>Hello, world!\nWhat's up?",
/* .content_parts = */ {},
/* .tool_calls = */ {},
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_thoughts {
"assistant",
"Hello, world!\nWhat's up?",
/* .content_parts = */ {},
/* .tool_calls = */ {},
/* .reasoning_content = */ "I'm thinking",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const std::vector<common_chat_tool_call> tool_calls {
{ "special_function", "{\"arg1\": 1}", /* .id = */ "" },
};
const std::vector<common_chat_tool_call> tool_calls_idx {
{ "special_function", "{\"arg1\": 1}", /* .id = */ "0" },
};
const std::vector<common_chat_tool_call> tool_calls_id {
{ "special_function", "{\"arg1\": 1}", /* .id = */ "123456789" },
};
const common_chat_msg message_assist_call {
"assistant",
"",
/* .content_parts = */ {},
tool_calls,
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_call_thoughts = {
"assistant",
/* .content = */ "",
/* .content_parts = */ {},
tool_calls,
/* .reasoning_content = */ "I'm\nthinking",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_call_thoughts_unparsed = {
"assistant",
/* .content = */ "<think>I'm\nthinking</think>",
/* .content_parts = */ {},
tool_calls,
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_call_id {
"assistant",
"",
/* .content_parts = */ {},
tool_calls_id,
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_call_idx {
"assistant",
"",
/* .content_parts = */ {},
tool_calls_idx,
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_call_python {
"assistant",
"",
/* .content_parts = */ {},
{ { "python", "{\"code\": \"print('hey')\"}", /* .id = */ "" } },
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
const common_chat_msg message_assist_call_code_interpreter {
"assistant",
"",
/* .content_parts = */ {},
{ { "code_interpreter", "{\"code\": \"print('hey')\"}", /* .id = */ "" } },
/* .reasoning_content = */ "",
/* .tool_name = */ "",
/* .tool_call_id = */ "",
};
static void test_msgs_oaicompat_json_conversion() {
std::vector<common_chat_msg> msgs{
message_user,
message_user_parts,
message_assist_call,
message_assist_call_thoughts,
message_assist_call_thoughts_unparsed,
message_assist_call_id,
message_assist_call_idx,
message_assist_call_python,
message_assist_call_code_interpreter,
};
for (const auto & msg : msgs) {
auto oai_json = common_chat_msgs_to_json_oaicompat<json>({msg});
auto msgs2 = common_chat_msgs_parse_oaicompat(oai_json);
assert_equals((size_t) 1, msgs2.size());
auto msg2 = msgs2[0];
assert_msg_equals(msg, msg2);
}
assert_equals(
std::string(
"[\n"
" {\n"
" \"role\": \"user\",\n"
" \"content\": [\n"
" {\n"
" \"type\": \"text\",\n"
" \"text\": \"Hey\"\n"
" },\n"
" {\n"
" \"type\": \"text\",\n"
" \"text\": \"there\"\n"
" }\n"
" ]\n"
" }\n"
"]"
),
common_chat_msgs_to_json_oaicompat<json>({message_user_parts}).dump(2));
assert_equals(
std::string(
"[\n"
" {\n"
" \"role\": \"assistant\",\n"
" \"content\": null,\n"
" \"tool_calls\": [\n"
" {\n"
" \"type\": \"function\",\n"
" \"function\": {\n"
" \"name\": \"python\",\n"
" \"arguments\": \"{\\\"code\\\": \\\"print('hey')\\\"}\"\n"
" }\n"
" }\n"
" ]\n"
" }\n"
"]"
),
common_chat_msgs_to_json_oaicompat<json>({message_assist_call_python}).dump(2));
}
static void test_tools_oaicompat_json_conversion() {
std::vector<common_chat_tool> tools{
special_function_tool,
python_tool,
code_interpreter_tool,
};
for (const auto & tool : tools) {
auto oai_json = common_chat_tools_to_json_oaicompat<json>({tool});
auto tools2 = common_chat_tools_parse_oaicompat(oai_json);
assert_equals((size_t) 1, tools2.size());
auto tool2 = tools2[0];
assert_equals(tool.name, tool2.name);
assert_equals(tool.description, tool2.description);
assert_equals(json::parse(tool.parameters).dump(2), json::parse(tool2.parameters).dump(2));
}
assert_equals(
std::string(
"[\n"
" {\n"
" \"type\": \"function\",\n"
" \"function\": {\n"
" \"name\": \"special_function\",\n"
" \"description\": \"I'm special\",\n"
" \"parameters\": {\n"
" \"type\": \"object\",\n"
" \"properties\": {\n"
" \"arg1\": {\n"
" \"type\": \"integer\",\n"
" \"description\": \"The arg.\"\n"
" }\n"
" },\n"
" \"required\": [\n"
" \"arg1\"\n"
" ]\n"
" }\n"
" }\n"
" }\n"
"]"
),
common_chat_tools_to_json_oaicompat<json>({special_function_tool}).dump(2));
}
static void test_template_output_parsers() {
common_chat_templates_inputs inputs_no_tools;
inputs_no_tools.messages = {message_user};
inputs_no_tools.extract_reasoning = false;
common_chat_templates_inputs inputs_no_tools_think;
inputs_no_tools_think.messages = {message_user};
inputs_no_tools_think.extract_reasoning = true;
common_chat_templates_inputs inputs_tools;
inputs_tools.messages = {message_user};
inputs_tools.tools = {special_function_tool};
inputs_tools.extract_reasoning = false;
common_chat_templates_inputs inputs_tools_think;
inputs_tools_think.messages = {message_user};
inputs_tools_think.tools = {special_function_tool};
inputs_tools_think.extract_reasoning = true;
common_chat_templates_inputs inputs_tools_builtin;
inputs_tools_builtin.messages = {message_user};
inputs_tools_builtin.tools = {python_tool};
inputs_tools_builtin.extract_reasoning = false;
{
// Not supported yet
auto tmpls = read_templates("models/templates/CohereForAI-c4ai-command-r-plus-tool_use.jinja");
assert_equals(COMMON_CHAT_FORMAT_GENERIC, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
}
{
auto tmpls = read_templates("models/templates/CohereForAI-c4ai-command-r7b-12-2024-tool_use.jinja");
std::vector<std::string> end_tokens{ "<|END_OF_TURN_TOKEN|>" };
assert_equals(COMMON_CHAT_FORMAT_COMMAND_R7B, common_chat_templates_apply(tmpls.get(), inputs_no_tools).format);
assert_equals(COMMON_CHAT_FORMAT_COMMAND_R7B, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
assert_equals(COMMON_CHAT_FORMAT_COMMAND_R7B_EXTRACT_REASONING, common_chat_templates_apply(tmpls.get(), inputs_tools_think).format);
assert_msg_equals(message_assist,
common_chat_parse(
"Hello, world!\nWhat's up?",
COMMON_CHAT_FORMAT_COMMAND_R7B));
assert_msg_equals(message_assist,
common_chat_parse(
"Hello, world!\nWhat's up?<|END_RESPONSE|>",
COMMON_CHAT_FORMAT_COMMAND_R7B));
assert_msg_equals(message_assist,
common_chat_parse(
"<|START_RESPONSE|>Hello, world!\nWhat's up?<|END_RESPONSE|>",
COMMON_CHAT_FORMAT_COMMAND_R7B));
assert_msg_equals(message_assist_thoughts_unparsed_r7b,
common_chat_parse(
"<|START_THINKING|>I'm thinking<|END_THINKING|>"
"<|START_RESPONSE|>Hello, world!\nWhat's up?<|END_RESPONSE|>",
COMMON_CHAT_FORMAT_COMMAND_R7B));
assert_msg_equals(message_assist_thoughts_unparsed_r7b,
common_chat_parse(
"<|START_THINKING|>I'm thinking<|END_THINKING|>"
"Hello, world!\nWhat's up?<|END_RESPONSE|>",
COMMON_CHAT_FORMAT_COMMAND_R7B));
assert_msg_equals(message_assist_thoughts,
common_chat_parse(
"<|START_THINKING|>I'm thinking<|END_THINKING|>"
"<|START_RESPONSE|>Hello, world!\nWhat's up?<|END_RESPONSE|>",
COMMON_CHAT_FORMAT_COMMAND_R7B_EXTRACT_REASONING));
test_templates(tmpls.get(), end_tokens, message_assist_call_idx, tools,
"<|START_THINKING|><|END_THINKING|>"
"<|START_ACTION|>[\n"
" {\"tool_call_id\": \"0\", \"tool_name\": \"special_function\", \"parameters\": {\"arg1\": 1}}\n"
"]<|END_ACTION|>");
test_templates(tmpls.get(), end_tokens, message_assist, tools,
"<|START_RESPONSE|>Hello, world!\n"
"What's up?<|END_RESPONSE|>",
/* expect_grammar_triggered= */ false);
}
{
auto tmpls = read_templates("models/templates/google-gemma-2-2b-it.jinja");
std::vector<std::string> end_tokens{ "<end_of_turn>" };
assert_equals(COMMON_CHAT_FORMAT_CONTENT_ONLY, common_chat_templates_apply(tmpls.get(), inputs_no_tools).format);
assert_equals(COMMON_CHAT_FORMAT_GENERIC, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
assert_equals(COMMON_CHAT_FORMAT_GENERIC,
common_chat_templates_apply(
read_templates("models/templates/microsoft-Phi-3.5-mini-instruct.jinja").get(),
inputs_tools)
.format);
// Generic tool calls doesn't generate / parse content-only messages symmetrically.
assert_msg_equals(message_assist,
common_chat_parse("{\n"
" \"response\": \"Hello, world!\\nWhat's up?\"\n"
"}",
common_chat_templates_apply(tmpls.get(), inputs_tools).format));
test_templates(tmpls.get(), end_tokens, message_assist_call_id, tools,
"{\n"
" \"tool_calls\": [\n"
" {\n"
" \"name\": \"special_function\",\n"
" \"arguments\": {\n"
" \"arg1\": 1\n"
" },\n"
" \"id\": \"123456789\"\n"
" }\n"
" ]\n"
"}");
}
{
auto tmpls = read_templates("models/templates/mistralai-Mistral-Nemo-Instruct-2407.jinja");
std::vector<std::string> end_tokens{ "</s>" };
assert_equals(COMMON_CHAT_FORMAT_MISTRAL_NEMO, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
test_templates(tmpls.get(), end_tokens, message_assist, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
test_templates(
tmpls.get(), end_tokens, message_assist_call_id, tools,
"[TOOL_CALLS][{\"name\": \"special_function\", \"arguments\": {\"arg1\": 1}, \"id\": \"123456789\"}]");
}
{
auto tmpls = read_templates("models/templates/NousResearch-Hermes-2-Pro-Llama-3-8B-tool_use.jinja");
std::vector<std::string> end_tokens{ "<|im_end|>" };
assert_equals(COMMON_CHAT_FORMAT_HERMES_2_PRO, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
assert_equals(
COMMON_CHAT_FORMAT_HERMES_2_PRO,
common_chat_templates_apply(
read_templates("models/templates/NousResearch-Hermes-3-Llama-3.1-8B-tool_use.jinja").get(),
inputs_tools)
.format);
assert_equals(
COMMON_CHAT_FORMAT_HERMES_2_PRO,
common_chat_templates_apply(
read_templates("models/templates/Qwen-Qwen2.5-7B-Instruct.jinja").get(),
inputs_tools)
.format);
test_templates(tmpls.get(), end_tokens, message_assist, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
test_templates(tmpls.get(), end_tokens, message_assist_call, tools,
"<tool_call>\n"
"{\"name\": \"special_function\", \"arguments\": {\"arg1\": 1}}\n"
"</tool_call>");
test_templates(tmpls.get(), end_tokens, message_assist_call_python, tools,
"<tool_call>\n"
"{\"name\": \"python\", \"arguments\": {\"code\": \"print('hey')\"}}\n"
"</tool_call>");
}
{
auto tmpls = read_templates("models/templates/meta-llama-Llama-3.1-8B-Instruct.jinja");
std::vector<std::string> end_tokens{ "<|eom_id|>", "<|eot_id|>" };
assert_equals(COMMON_CHAT_FORMAT_LLAMA_3_X, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
assert_equals(COMMON_CHAT_FORMAT_LLAMA_3_X_WITH_BUILTIN_TOOLS,
common_chat_templates_apply(tmpls.get(), inputs_tools_builtin).format);
assert_equals(COMMON_CHAT_FORMAT_LLAMA_3_X_WITH_BUILTIN_TOOLS,
common_chat_templates_apply(
read_templates("models/templates/meta-llama-Llama-3.3-70B-Instruct.jinja").get(),
inputs_tools_builtin)
.format);
// test_templates(tmpls.get(), end_tokens, message_assist, tools, R"(?)", /* expect_grammar_triggered= */ false);
test_templates(tmpls.get(), end_tokens, message_assist_call_code_interpreter, llama_3_1_tools,
"<|python_tag|>code_interpreter.call(code=\"print('hey')\")");
test_templates(tmpls.get(), end_tokens, message_assist_call_python, tools,
"<|python_tag|>python.call(code=\"print('hey')\")");
test_templates(tmpls.get(), end_tokens, message_assist_call, tools,
"{\"name\": \"special_function\", \"parameters\": {\"arg1\": 1}}");
}
{
auto tmpls = read_templates("models/templates/meta-llama-Llama-3.2-3B-Instruct.jinja");
std::vector<std::string> end_tokens{ "<|eom_id|>", "<|eot_id|>" };
assert_equals(COMMON_CHAT_FORMAT_LLAMA_3_X, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
test_templates(tmpls.get(), end_tokens, message_assist, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
test_templates(tmpls.get(), end_tokens, message_assist_call, tools,
"{\"name\": \"special_function\", \"parameters\": {\"arg1\": 1}}");
}
{
auto tmpls = read_templates("models/templates/meetkai-functionary-medium-v3.1.jinja");
std::vector<std::string> end_tokens{ "<|eom_id|>", "<|eot_id|>" };
assert_equals(COMMON_CHAT_FORMAT_FUNCTIONARY_V3_1_LLAMA_3_1,
common_chat_templates_apply(tmpls.get(), inputs_tools).format);
test_templates(tmpls.get(), end_tokens, message_assist, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
test_templates(tmpls.get(), end_tokens, message_assist_call, tools,
"<function=special_function>{\"arg1\": 1}</function>");
}
{
auto tmpls = read_templates("models/templates/meetkai-functionary-medium-v3.2.jinja");
std::vector<std::string> end_tokens{ "<|eom_id|>", "<|eot_id|>" };
assert_equals(COMMON_CHAT_FORMAT_FUNCTIONARY_V3_2, common_chat_templates_apply(tmpls.get(), inputs_no_tools).format);
assert_equals(COMMON_CHAT_FORMAT_FUNCTIONARY_V3_2, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
test_templates(tmpls.get(), end_tokens, message_assist, {},
"all\n"
"Hello, world!\n"
"What's up?",
/* expect_grammar_triggered= */ false);
test_templates(tmpls.get(), end_tokens, message_assist_call, tools,
"special_function\n"
"{\"arg1\": 1}");
}
{
auto tmpls = read_templates("models/templates/fireworks-ai-llama-3-firefunction-v2.jinja");
std::vector<std::string> end_tokens{ "<|eot_id|>" };
assert_equals(COMMON_CHAT_FORMAT_FIREFUNCTION_V2, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
test_templates(tmpls.get(), end_tokens, message_assist, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
test_templates(tmpls.get(), end_tokens, message_assist_call, tools,
" functools[{\"name\": \"special_function\", \"arguments\": {\"arg1\": 1}}]");
}
{
// Original DeepSeek R1 template. Leaves <｜tool▁calls▁begin｜> and others unclosed. Our logic fixes the prompt.
auto tmpls = read_templates("models/templates/deepseek-ai-DeepSeek-R1-Distill-Llama-8B.jinja");
std::vector<std::string> end_tokens{ "<｜end▁of▁sentence｜>" };
assert_equals(COMMON_CHAT_FORMAT_DEEPSEEK_R1, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
assert_equals(COMMON_CHAT_FORMAT_DEEPSEEK_R1_EXTRACT_REASONING, common_chat_templates_apply(tmpls.get(), inputs_tools_think).format);
test_templates(tmpls.get(), end_tokens, message_assist, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
test_templates(tmpls.get(), end_tokens, message_assist_thoughts, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
assert_msg_equals(message_assist_thoughts_unparsed_think,
common_chat_parse("<think>I'm thinking</think>Hello, world!\nWhat's up?",
COMMON_CHAT_FORMAT_DEEPSEEK_R1));
assert_msg_equals(message_assist_thoughts,
common_chat_parse("<think>I'm thinking</think>Hello, world!\nWhat's up?",
COMMON_CHAT_FORMAT_DEEPSEEK_R1_EXTRACT_REASONING));
assert_msg_equals(message_assist_thoughts,
// Latest template update (ast of 20250209) adds a trailing <think>\n if add_generation_prompt is true.
common_chat_parse("I'm thinking</think>Hello, world!\nWhat's up?",
COMMON_CHAT_FORMAT_DEEPSEEK_R1_EXTRACT_REASONING));
// test_templates(tmpls.get(), end_tokens, message_assist_call, tools,
// "<｜tool▁calls▁begin｜><｜tool▁call▁begin｜>function<｜tool▁sep｜>special_function\n"
// "```json\n"
// "{\"arg1\": 1}\n"
// // Look what's not here: <｜tool▁calls▁end｜> (also missing the <｜end▁of▁sentence｜>, but that is removed lazily by the test's delta logic)
// "```<｜tool▁call▁end｜>",
// /* expect_grammar_triggered= */ true,
// /* test_grammar_if_triggered= */ false);
}
{
// Replacement DeepSeek R1 template. Makes the Distill Qwen 7B/32B models happy to call tools and all.
auto tmpls = read_templates("models/templates/llama-cpp-deepseek-r1.jinja");
std::vector<std::string> end_tokens{ "<｜end▁of▁sentence｜>" };
assert_equals(COMMON_CHAT_FORMAT_DEEPSEEK_R1, common_chat_templates_apply(tmpls.get(), inputs_tools).format);
assert_equals(COMMON_CHAT_FORMAT_DEEPSEEK_R1_EXTRACT_REASONING, common_chat_templates_apply(tmpls.get(), inputs_tools_think).format);
test_templates(tmpls.get(), end_tokens, message_assist, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
test_templates(tmpls.get(), end_tokens, message_assist_thoughts, tools, "Hello, world!\nWhat's up?", /* expect_grammar_triggered= */ false);
assert_msg_equals(message_assist_thoughts_unparsed_think,
common_chat_parse("<think>I'm thinking</think>Hello, world!\nWhat's up?",
COMMON_CHAT_FORMAT_DEEPSEEK_R1));
assert_msg_equals(message_assist_thoughts,
common_chat_parse("<think>I'm thinking</think>Hello, world!\nWhat's up?",
COMMON_CHAT_FORMAT_DEEPSEEK_R1_EXTRACT_REASONING));
assert_msg_equals(message_assist_call_thoughts_unparsed,
common_chat_parse(
"<think>I'm\nthinking</think>\n\n"
"<｜tool▁calls▁begin｜><｜tool▁call▁begin｜>function<｜tool▁sep｜>special_function\n"
"```json\n"
"{\"arg1\": 1}\n"
"```<｜tool▁call▁end｜><｜tool▁calls▁end｜>",
COMMON_CHAT_FORMAT_DEEPSEEK_R1));
assert_msg_equals(message_assist_call_thoughts,
common_chat_parse(
"<think>I'm\nthinking</think>\n\n"
"<｜tool▁calls▁begin｜><｜tool▁call▁begin｜>function<｜tool▁sep｜>special_function\n"
"```json\n"
"{\"arg1\": 1}\n"
"```<｜tool▁call▁end｜><｜tool▁calls▁end｜>",
COMMON_CHAT_FORMAT_DEEPSEEK_R1_EXTRACT_REASONING));
test_templates(tmpls.get(), end_tokens, message_assist_call, tools,
"<｜tool▁calls▁begin｜><｜tool▁call▁begin｜>function<｜tool▁sep｜>special_function\n"
"```json\n"
"{\"arg1\": 1}\n"
"```<｜tool▁call▁end｜><｜tool▁calls▁end｜>");
}
}
int main(int argc, char ** argv) {
try {
#ifndef _WIN32
if (argc > 1) {
common_chat_templates_inputs inputs;
common_chat_msg msg;
msg.role = "user";
msg.content = "Hey";
inputs.messages = {msg};
inputs.tools = { special_function_tool };
std::cout << "| Template | Format |\n";
std::cout << "|----------|--------|\n";
for (int i = 1; i < argc; i++) {
try {
std::string path = argv[i];
if (path.rfind(".jinja") != path.size() - 6) {
std::cerr << "Skipping non-jinja file: " << path << '\n';
continue;
}
auto tmpls = read_templates(path);
auto parts = string_split(path, "/");
auto name = parts[parts.size() - 1];
auto format = common_chat_format_name(common_chat_templates_apply(tmpls.get(), inputs).format);
std::cout << "| " << name << " | " << format << " |\n";
} catch (const std::exception & e) {
std::cerr << "Failed to process " << argv[i] << ": " << e.what() << '\n';
}
}
} else
#endif
{
test_msgs_oaicompat_json_conversion();
test_tools_oaicompat_json_conversion();
test_template_output_parsers();
std::cout << "\n[chat] All tests passed!" << '\n';
}
return 0;
} catch (const std::exception & e) {
std::cerr << "Error: " << e.what() << '\n';
return 1;
}
}

1338
tests/test-gguf.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

13
tests/test-grammar-integration.cpp
View File

@@ -13,7 +13,7 @@
using json = nlohmann::ordered_json;
static llama_grammar * build_grammar(const std::string & grammar_str) {
return llama_grammar_init_impl(nullptr, grammar_str.c_str(), "root");
return llama_grammar_init_impl(nullptr, grammar_str.c_str(), "root", false, nullptr, 0, nullptr, 0);
}
static bool test_build_grammar_fails(const std::string & grammar_str) {
@@ -32,13 +32,10 @@ static bool test_build_grammar_fails(const std::string & grammar_str) {
static bool match_string(const std::string & input, llama_grammar * grammar) {
const auto cpts = unicode_cpts_from_utf8(input);
const llama_grammar_rules & rules = llama_grammar_get_rules (grammar);
llama_grammar_stacks & stacks_cur = llama_grammar_get_stacks(grammar);
auto & stacks_cur = llama_grammar_get_stacks(grammar);
for (const auto & cpt : cpts) {
const llama_grammar_stacks stacks_prev = llama_grammar_get_stacks(grammar); // copy
llama_grammar_accept(rules, stacks_prev, cpt, stacks_cur);
llama_grammar_accept(grammar, cpt);
if (stacks_cur.empty()) {
// no stacks means that the grammar failed to match at this point
@@ -63,7 +60,7 @@ static void test(const std::string & test_desc, const std::string & grammar_str,
auto * grammar = build_grammar(grammar_str);
// Save the original grammar stacks so that we can reset after every new string we want to test
const llama_grammar_stacks stacks_org = llama_grammar_get_stacks(grammar);
const llama_grammar_stacks stacks_org = llama_grammar_get_stacks(grammar); // copy
llama_grammar_stacks & stacks_cur = llama_grammar_get_stacks(grammar);
@@ -132,7 +129,7 @@ static void test_grammar(const std::string & test_desc, const std::string & gram
test(test_desc + ". Grammar: " + grammar_str, grammar_str, passing_strings, failing_strings);
}
static void test_schema(const std::string & test_desc, const std::string & schema_str, const std::vector<std::string> & passing_strings, const std::vector<std::string> & failing_strings) {
test(test_desc + ". Schema: " + schema_str, json_schema_to_grammar(json::parse(schema_str)), passing_strings, failing_strings);
test(test_desc + ". Schema: " + schema_str, json_schema_to_grammar(json::parse(schema_str), true), passing_strings, failing_strings);
}
static void test_simple_grammar() {

1140
tests/test-grammar-llguidance.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

2
tests/test-json-schema-to-grammar.cpp
View File

@@ -1246,7 +1246,7 @@ int main() {
test_all("C++", [](const TestCase & tc) {
try {
tc.verify(json_schema_to_grammar(nlohmann::ordered_json::parse(tc.schema)));
tc.verify(json_schema_to_grammar(nlohmann::ordered_json::parse(tc.schema), true));
tc.verify_status(SUCCESS);
} catch (const std::runtime_error & ex) {
fprintf(stderr, "Error: %s\n", ex.what());

6
tests/test-llama-grammar.cpp
View File

@@ -113,12 +113,10 @@ int main()
}
}
llama_grammar * grammar = NULL;
std::vector<const llama_grammar_element *> grammar_rules(parsed_grammar.c_rules());
grammar = llama_grammar_init_impl(nullptr, grammar_rules.data(), grammar_rules.size(), parsed_grammar.symbol_ids.at("root"));
if (grammar == nullptr)
{
llama_grammar * grammar = llama_grammar_init_impl(nullptr, grammar_rules.data(), grammar_rules.size(), parsed_grammar.symbol_ids.at("root"));
if (grammar == nullptr) {
throw std::runtime_error("Failed to initialize llama_grammar");
}

11
tests/test-lora-conversion-inference.sh
View File

@@ -10,11 +10,16 @@ declare -a params=(
MODELS_REPO=lora-tests
MODELS_REPO_URL=https://huggingface.co/ggml-org/$MODELS_REPO
COMMIT=c26d5fb85b4070a9e9c4e65d132c783b98086890
# Clone the Hugging Face repository if the directory does not exist
if [ ! -d "$MODELS_REPO" ]; then
echo "Cloning the Hugging Face repository..."
git clone $MODELS_REPO_URL --depth 1
cd $MODELS_REPO
git fetch --depth=1 origin $COMMIT
git reset --hard $COMMIT
cd -
else
echo "Repository already exists. Skipping clone."
fi
@@ -75,18 +80,18 @@ run_conversion_and_inference_lora() {
# Run inference
echo -e "\n\n---------------------------\n\n"
echo "Running llama-cli without lora for $model_name with hidden_size $hidden_size..."
OUTPUT_BASE=$(./llama-cli -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32.gguf \
OUTPUT_BASE=$(./llama-cli -no-cnv -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32.gguf \
-p "$EXPECTED_BASE_FIRST_WORD" -n 50 --seed 42 --temp 0)
echo -e "\n\n---------------------------\n\n"
echo "Running llama-cli with hot lora for $model_name with hidden_size $hidden_size..."
OUTPUT_LORA_HOT=$(./llama-cli -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32.gguf \
OUTPUT_LORA_HOT=$(./llama-cli -no-cnv -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32.gguf \
--lora $MODELS_REPO/$model_name/hidden_size=$hidden_size/lora/Lora-F32-LoRA.gguf \
-p "$EXPECTED_LORA_FIRST_WORD" -n 50 --seed 42 --temp 0)
echo -e "\n\n---------------------------\n\n"
echo "Running llama-cli with merged lora for $model_name with hidden_size $hidden_size..."
OUTPUT_LORA_MERGED=$(./llama-cli -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32-lora-merged.gguf \
OUTPUT_LORA_MERGED=$(./llama-cli -no-cnv -m $MODELS_REPO/$model_name/hidden_size=$hidden_size/base/Base-F32-lora-merged.gguf \
-p "$EXPECTED_LORA_FIRST_WORD" -n 50 --seed 42 --temp 0)
# Remove any initial white space

2
tests/test-model-load-cancel.cpp
View File

@@ -21,7 +21,7 @@ int main(int argc, char *argv[] ) {
(void) ctx;
return progress > 0.50;
};
auto * model = llama_load_model_from_file(model_path, params);
auto * model = llama_model_load_from_file(model_path, params);
llama_backend_free();
return model == nullptr ? EXIT_SUCCESS : EXIT_FAILURE;
}

81
tests/test-rope.cpp
View File

@@ -138,7 +138,7 @@ int main(int /*argc*/, const char ** /*argv*/) {
struct ggml_tensor * x;
// rope f32
for (int m = 0; m < 3; ++m) {
for (int m = 0; m < 5; ++m) {
const int ndims = 4;
const int64_t n_rot = 128;
@@ -147,28 +147,69 @@ int main(int /*argc*/, const char ** /*argv*/) {
const int n_past_0 = 100;
const int n_past_2 = 33;
struct ggml_tensor * p0 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2]);
struct ggml_tensor * p1 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2]);
struct ggml_tensor * p2 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2]);
for (int i = 0; i < ne[2]; ++i) {
((int32_t *) p0->data)[i] = n_past_0 + i;
((int32_t *) p1->data)[i] = n_past_2 - n_past_0;
((int32_t *) p2->data)[i] = n_past_2 + i;
}
// test mode 0, 2, 4 (standard, GPT-NeoX, GLM)
const int mode = m == 0 ? 0 : m == 1 ? 2 : 4;
struct ggml_tensor * r0;
struct ggml_tensor * r1;
struct ggml_tensor * r2;
x = get_random_tensor_f32(ctx0, ndims, ne, -1.0f, 1.0f);
int mode = -1;
// 100, 101, 102, ..., 172
struct ggml_tensor * r0 = ggml_rope(ctx0, x, p0, n_rot, mode);
// -67, -67, -67, ..., -67
struct ggml_tensor * r1 = ggml_rope(ctx0, r0, p1, n_rot, mode); // "context swap", i.e. forget n_past_0 - n_past_2 tokens
if (m < 3) {
struct ggml_tensor * p0 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2]);
struct ggml_tensor * p1 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2]);
struct ggml_tensor * p2 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2]);
// 33, 34, 35, ..., 105
struct ggml_tensor * r2 = ggml_rope(ctx0, x, p2, n_rot, mode);
for (int i = 0; i < ne[2]; ++i) {
((int32_t *) p0->data)[i] = n_past_0 + i;
((int32_t *) p1->data)[i] = n_past_2 - n_past_0;
((int32_t *) p2->data)[i] = n_past_2 + i;
}
// test mode 0, 2, 4 (standard, GPT-NeoX, GLM)
mode = m == 0 ? 0 : m == 1 ? 2 : 4;
// 100, 101, 102, ..., 172
r0 = ggml_rope(ctx0, x, p0, n_rot, mode);
// -67, -67, -67, ..., -67
r1 = ggml_rope(ctx0, r0, p1, n_rot, mode); // "context swap", i.e. forget n_past_0 - n_past_2 tokens
// 33, 34, 35, ..., 105
r2 = ggml_rope(ctx0, x, p2, n_rot, mode);
} else {
// testing multi-dimension rope position embedding mode
struct ggml_tensor * p0 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2] * 4);
struct ggml_tensor * p1 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2] * 4);
struct ggml_tensor * p2 = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ne[2] * 4);
int sections[4] = {16, 24, 24, 0};
mode = (m == 3) ? GGML_ROPE_TYPE_MROPE : GGML_ROPE_TYPE_VISION;
for (int i = 0; i < ne[2]; ++i) {
for (int j = 0; j < 4; ++j) {
((int32_t *) p0->data)[i + ne[2] * j] = n_past_0 + i + j;
((int32_t *) p1->data)[i + ne[2] * j] = n_past_2 - n_past_0;
((int32_t *) p2->data)[i + ne[2] * j] = n_past_2 + i + j;
}
}
// [[100, 101, 102, ..., 172],
// [101, 102, 103, ..., 173],
// [102, 103, 104, ..., 174]]
r0 = ggml_rope_multi(
ctx0, x, p0, nullptr,
n_rot, sections, mode, 32768, 1000000, 1, 0, 1, 32, 1);
// [[-67, -67, -67, ..., -67]
// [-67, -67, -67, ..., -67]
// [-67, -67, -67, ..., -67]]
r1 = ggml_rope_multi(
ctx0, r0, p1, nullptr,
n_rot, sections, mode, 32768, 1000000, 1, 0, 1, 32, 1);
// [[33, 34, 35, ..., 105]
// [34, 35, 36, ..., 106]
// [35, 36, 37, ..., 107]]
r2 = ggml_rope_multi(
ctx0, x, p2, nullptr,
n_rot, sections, mode, 32768, 1000000, 1, 0, 1, 32, 1);
}
ggml_cgraph * gf = ggml_new_graph(ctx0);

20
tests/test-sampling.cpp
View File

@@ -144,8 +144,7 @@ static void test_penalties(
sampler_tester tester(probs, probs_expected);
const size_t n_vocab = probs.size();
auto * sampler = llama_sampler_init_penalties(n_vocab, LLAMA_TOKEN_NULL, LLAMA_TOKEN_NULL, last_tokens.size(), repeat_penalty, alpha_frequency, alpha_presence, false, false);
auto * sampler = llama_sampler_init_penalties(last_tokens.size(), repeat_penalty, alpha_frequency, alpha_presence);
for (size_t i = 0; i < last_tokens.size(); i++) {
llama_sampler_accept(sampler, last_tokens[i]);
@@ -182,6 +181,17 @@ static void test_dry(
tester.check();
}
static void test_top_n_sigma(const std::vector<float> & probs, const std::vector<float> & probs_expected, int n) {
sampler_tester tester(probs, probs_expected);
DUMP(&tester.cur_p);
tester.apply(llama_sampler_init_top_n_sigma(n));
tester.apply(llama_sampler_init_dist (0));
DUMP(&tester.cur_p);
tester.check();
}
static void test_sampler_queue(const size_t n_vocab, const std::string & samplers_sequence, const int top_k, const float top_p, const float min_p
) {
sampler_tester tester(n_vocab);
@@ -284,7 +294,7 @@ static void test_perf() {
data.reserve(n_vocab);
for (int i = 0; i < n_vocab; i++) {
const float logit = 2.0f*((float)(rand())/RAND_MAX - 0.5f);
const float logit = 2.0f*((double)(rand())/RAND_MAX - 0.5);
data.emplace_back(llama_token_data{i, logit, 0.0f});
}
@@ -349,6 +359,10 @@ int main(void) {
test_dry({0.2f, 0.2f, 0.2f, 0.2f, 0.2f}, {0, 1, 2, 0, 1}, {0.241818f, 0.241818f, 0.241818f, 0.241818f, 0.032727f}, 2.0f, 1.1f, 2, 5, {});
test_dry({0.2f, 0.2f, 0.2f, 0.2f, 0.2f}, {0, 1, 2, 3, 4, 0, 1}, {0.2f, 0.2f, 0.2f, 0.2f, 0.2f}, 1.0f, 1.1f, 4, 7, {});
test_top_n_sigma({0.1f, 0.2f, 0.3f, 0.4f}, {0.571429f, 0.428571f, 0.0f, 0.0f}, 1.00f);
test_top_n_sigma({0.1f, 0.2f, 0.3f, 0.4f}, {1.0f, 0.0f, 0.0f, 0.0f}, 0.00f);
test_top_n_sigma({0.1f, 0.2f, 0.3f, 0.4f}, {0.4f, 0.3f, 0.2f, 0.1f}, 3.00f);
test_sampler_queue(10000, "k", 10000, 1.0f, 1.0f);
test_sampler_queue(10000, "k", 1, 1.0f, 1.0f);
test_sampler_queue(10000, "p", 10000, 1.0f, 1.0f);

8
tests/test-tokenizer-0.cpp
View File

@@ -152,7 +152,7 @@ int main(int argc, char **argv) {
mparams.vocab_only = true;
model = llama_load_model_from_file(fname.c_str(), mparams);
model = llama_model_load_from_file(fname.c_str(), mparams);
if (model == NULL) {
fprintf(stderr, "%s: error: failed to load vocab '%s'\n", __func__, fname.c_str());
@@ -161,11 +161,11 @@ int main(int argc, char **argv) {
auto cparams = llama_context_default_params();
ctx = llama_new_context_with_model(model, cparams);
ctx = llama_init_from_model(model, cparams);
if (ctx == NULL) {
fprintf(stderr, "%s: error: failed to load vocab '%s'\n", __func__, fname.c_str());
llama_free_model(model);
llama_model_free(model);
return 1;
}
}
@@ -300,7 +300,7 @@ int main(int argc, char **argv) {
fprintf(stderr, "%s : tokens written to '%s'\n", __func__, (fname_text + ".tokcpp").c_str());
}
llama_free_model(model);
llama_model_free(model);
llama_free(ctx);
llama_backend_free();

16
tests/test-tokenizer-1-bpe.cpp
View File

@@ -46,7 +46,7 @@ int main(int argc, char **argv) {
mparams.vocab_only = true;
model = llama_load_model_from_file(fname.c_str(), mparams);
model = llama_model_load_from_file(fname.c_str(), mparams);
if (model == NULL) {
fprintf(stderr, "%s: error: failed to load vocab '%s'\n", __func__, fname.c_str());
@@ -55,17 +55,19 @@ int main(int argc, char **argv) {
auto cparams = llama_context_default_params();
ctx = llama_new_context_with_model(model, cparams);
ctx = llama_init_from_model(model, cparams);
if (ctx == NULL) {
fprintf(stderr, "%s: error: failed to load vocab '%s'\n", __func__, fname.c_str());
llama_free_model(model);
llama_model_free(model);
return 1;
}
}
//GGML_ASSERT(llama_vocab_type(model) == LLAMA_VOCAB_TYPE_BPE);
if (llama_vocab_type(model) != LLAMA_VOCAB_TYPE_BPE) {
const llama_vocab * vocab = llama_model_get_vocab(model);
//GGML_ASSERT(llama_vocab_type(vocab) == LLAMA_VOCAB_TYPE_BPE);
if (llama_vocab_type(vocab) != LLAMA_VOCAB_TYPE_BPE) {
return 99;
}
@@ -75,7 +77,7 @@ int main(int argc, char **argv) {
atexit([]() { console::cleanup(); });
#endif
const int n_vocab = llama_n_vocab(model);
const int n_vocab = llama_vocab_n_tokens(vocab);
for (int i = 0; i < n_vocab; ++i) {
std::string str = common_detokenize(ctx, std::vector<int>(1, i));
@@ -143,7 +145,7 @@ int main(int argc, char **argv) {
}
}
llama_free_model(model);
llama_model_free(model);
llama_free(ctx);
llama_backend_free();

14
tests/test-tokenizer-1-spm.cpp
View File

@@ -34,7 +34,7 @@ int main(int argc, char ** argv) {
mparams.vocab_only = true;
model = llama_load_model_from_file(fname.c_str(), mparams);
model = llama_model_load_from_file(fname.c_str(), mparams);
if (model == NULL) {
fprintf(stderr, "%s: error: failed to load vocab '%s'\n", __func__, fname.c_str());
@@ -43,17 +43,19 @@ int main(int argc, char ** argv) {
auto cparams = llama_context_default_params();
ctx = llama_new_context_with_model(model, cparams);
ctx = llama_init_from_model(model, cparams);
if (ctx == NULL) {
fprintf(stderr, "%s: error: failed to load vocab '%s'\n", __func__, fname.c_str());
llama_free_model(model);
llama_model_free(model);
return 1;
}
}
const llama_vocab * vocab = llama_model_get_vocab(model);
//GGML_ASSERT(llama_vocab_type(model) == LLAMA_VOCAB_TYPE_SPM);
if (llama_vocab_type(model) != LLAMA_VOCAB_TYPE_SPM) {
if (llama_vocab_type(vocab) != LLAMA_VOCAB_TYPE_SPM) {
return 99;
}
@@ -63,7 +65,7 @@ int main(int argc, char ** argv) {
atexit([]() { console::cleanup(); });
#endif
const int n_vocab = llama_n_vocab(model);
const int n_vocab = llama_vocab_n_tokens(vocab);
for (int i = 0; i < n_vocab; ++i) {
std::string str = common_detokenize(ctx, std::vector<int>(1, i), true);
@@ -113,7 +115,7 @@ int main(int argc, char ** argv) {
}
}
llama_free_model(model);
llama_model_free(model);
llama_free(ctx);
llama_backend_free();

4
tests/test-tokenizer-random.py
View File

@@ -76,7 +76,7 @@ class LibLlamaModel:
self.ffi = libllama.ffi
if isinstance(mparams, dict):
mparams = libllama.model_default_params(**mparams)
self.model = self.lib.llama_load_model_from_file(path_model.encode(), mparams)
self.model = self.lib.llama_model_load_from_file(path_model.encode(), mparams)
if not self.model:
raise RuntimeError("error: failed to load model '%s'" % path_model)
if isinstance(cparams, dict):
@@ -92,7 +92,7 @@ class LibLlamaModel:
if self.ctx:
self.lib.llama_free(self.ctx)
if self.model:
self.lib.llama_free_model(self.model)
self.lib.llama_model_free(self.model)
self.ctx = None
self.model = None
self.lib = None