llama.cpp/examples/openai/prompting.py

596 lines
26 KiB
Python
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

from abc import ABC, abstractmethod
from enum import Enum
from functools import wraps
import jinja2
import json
from pathlib import Path
import random
import re
import sys
from typing import Any, Dict, Literal, Optional, Tuple, Callable, Union
from pydantic import BaseModel
from typeguard import typechecked
from examples.json_schema_to_grammar import SchemaConverter
from examples.openai.api import Tool, Message, FunctionCall, ToolCall
from examples.openai.gguf_kvs import GGUFKeyValues, Keys
from examples.openai.ts_converter import SchemaToTypeScriptConverter
@typechecked
def raise_exception(msg: str):
raise Exception(msg)
@typechecked
class ChatTemplate(BaseModel):
template: str
@property
def tool_style(self) -> 'ToolsPromptStyle':
return self._tool_style
def __init__(self, template: str, eos_token: str, bos_token: str):
super().__init__(template=template
)
env = jinja2.Environment(loader=jinja2.BaseLoader(), trim_blocks=True, lstrip_blocks=True)
self._template = env.from_string(template)
self._eos_token = eos_token
self._bos_token = bos_token
self._strict_user_assistant_alternation = "{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception" in template
if "<|recipient|>' + tool_call['function']['name']" in template:
self._tool_style = ToolsPromptStyle.TYPESCRIPT_FUNCTIONARY_V2
else:
self._tool_style = ToolsPromptStyle.TOOLS_BESPOKE
# self._tool_style = ToolsPromptStyle.TOOLS_LONG
# TODO: Test whether the template supports formatting tool_calls
delimiter = '<%$[SAMPLE]$%>'
user_msg = Message(role="user", content="Hey")
empty_prompt = self.render([user_msg], add_generation_prompt=True).strip()
planted_prompt = self.render([user_msg, Message(role="assistant", content=delimiter)], add_generation_prompt=False).strip()
assert planted_prompt.startswith(empty_prompt), f"Planted prompt does not start with empty prompt: {planted_prompt} vs {empty_prompt}"
[prefix, suffix] = planted_prompt[len(empty_prompt):].split(delimiter)
sys.stderr.write(f"\n# prefix={prefix}\n# suffix={suffix}\n\n")
self._prefix = prefix
self._suffix = suffix
def strip_suffix(self, s: str) -> str:
if s.endswith(self._suffix):
return s[:-len(self._suffix)]
else:
sys.stderr.write(f"Expected suffix ({self._suffix}) not found: {s}\n")
return s
def __str__(self):
return f"ChatTemplate(template={self.template}, eos_token={self._eos_token}, bos_token={self._bos_token})"
def add_system_prompt(self, messages: list[Message], system_prompt: Message) -> list[Message]:
assert system_prompt.role == "system"
# TODO: add to last system message, or create a new one just before the last user message
system_message = next(((i, m) for i, m in enumerate(messages) if m.role == "system"), None)
if system_message is not None:
(i, m) = system_message
return messages[:i] + [Message(role="system", content=system_prompt.content + '\n' + m.content)] + messages[i+1:]
else:
return [system_prompt] + messages
@staticmethod
def from_gguf(metadata: GGUFKeyValues):
tokens = metadata[Keys.Tokenizer.LIST]
return ChatTemplate(
template = metadata[Keys.Tokenizer.CHAT_TEMPLATE],
bos_token = tokens[metadata[Keys.Tokenizer.BOS_ID]],
eos_token = tokens[metadata[Keys.Tokenizer.EOS_ID]])
def render(self, messages: list[Message], add_generation_prompt: bool, omit_bos: bool = False):
if self._strict_user_assistant_alternation and any(m.role not in ('user', 'assistant') for m in messages):
new_messages=[]
i = 0
n = len(messages)
while i < n:
if messages[i].role == 'system':
assert messages[i+1].role == 'user'
new_messages.append(Message(
role="user",
content=f'[SYS]{messages[i].content}[/SYS]\n{messages[i+1].content}'
))
i += 2
elif messages[i].role == 'assistant' and messages[i].tool_calls and messages[i].content:
tc = '\n'.join(f'<tool_call>{json.dumps(tc.model_dump())}</tool_call>' for tc in messages[i].tool_calls)
new_messages.append(Message(
role="assistant",
content=f'{messages[i].content}\n{tc}'
))
i += 1
else:
new_messages.append(messages[i])
i += 1
# print(f'new_messages={json.dumps(new_messages, indent=2)}')
messages = new_messages
# print(f'messages={messages}')
result = self._template.render(
messages=messages,
eos_token=self._eos_token,
bos_token='' if omit_bos else self._bos_token,
raise_exception=raise_exception,
add_generation_prompt=add_generation_prompt,
)
sys.stderr.write(f'\n# RENDERED:\n\n{result}\n\n')
return result
# While the API will be usable with a generic tools usage like OpenAI,
# (see https://cookbook.openai.com/examples/how_to_call_functions_with_chat_models),
# each model may need specific prompting (and/or constrained output,
# especially for models not fine-tuned for tool usage / function calling).
class ToolsPromptStyle(Enum):
# Short prompt w/ <tools>schemas</tools>, <tool_call>...</tool_call> output
TOOLS_SHORT = 1
# Longer prompt w/ <tools>schemas</tools>, <tool_call>...</tool_call> output
TOOLS_LONG = 2
# Bespoke constrained output format that favours thought and reasoning
# while allowing unambiguous parsing of parallel tool calling.
TOOLS_BESPOKE = 3
# Large prompt for https://huggingface.co/NousResearch/Hermes-2-Pro-Mistral-7B
# <tool_call>...</tool_call> output
# Requires:
# - git clone https://github.com/NousResearch/Hermes-Function-Calling examples/openai/hermes_function_calling
# - Set large context length as their prompts are super long
TOOLS_HERMES_2_PRO = 4
# Seems to want to escape underscores in tool names and in the <tool\_call>...</tool\_call> tags
TOOLS_MISTRAL = 5
# Short prompt w/ TypeScript definitions for https://github.com/MeetKai/functionary
# https://github.com/MeetKai/functionary/blob/main/functionary/prompt_template/prompt_template_v2.py
# Note: see this prior attempt to support Functionary: https://github.com/ggerganov/llama.cpp/pull/5695
TYPESCRIPT_FUNCTIONARY_V2 = 6
class ChatHandlerArgs(BaseModel):
chat_template: ChatTemplate
response_schema: Optional[dict] = None
tools: Optional[list[Tool]] = None
class ChatHandler(ABC):
def __init__(self, args: ChatHandlerArgs):
self.args = args
self.output_format_prompt: Optional[Message] = None
self.grammar: Optional[str] = None
@abstractmethod
def parse(self, s: str) -> Optional[Message]:
raise NotImplementedError()
class NoToolsChatHandler(ChatHandler):
def __init__(self, args: ChatHandlerArgs):
super().__init__(args)
assert not args.tools
if args.response_schema:
self.output_format_prompt = Message(
role="system",
content=_please_respond_with_schema(args.response_schema)
)
converter = SchemaConverter(prop_order={}, allow_fetch=False, dotall=False, raw_pattern=False)
self.grammar = converter.visit(args.response_schema, '')
else:
self.output_format_prompt = None
self.grammar = None
@typechecked
def parse(self, s: str) -> Optional[Message]:
return Message(role="assistant", content=s)
class ToolCallTagsChatHandler(ChatHandler):
def __init__(self, args: ChatHandlerArgs, escapes_underscores: bool, allow_parallel_calls: bool):
super().__init__(args)
converter = SchemaConverter(prop_order={}, allow_fetch=False, dotall=False, raw_pattern=False)
tool_rules = [
converter.visit(
dict(
type="object",
properties=dict(
name=dict(type="string", pattern='^' + tool.function.name.replace('_', f'\\?_') + '$') if escapes_underscores \
else dict(const=tool.function.name),
arguments=tool.function.parameters,
),
required=['name', 'arguments']
),
f'{tool.function.name}-tool-call'
)
for tool in self.args.tools
]
def format_literal(s: str) -> str:
if escapes_underscores:
return ' "\\\\"? "_" '.join((converter._format_literal(part) for part in s.split('_')))
else:
return converter._format_literal(s)
tool_call_rule = converter._add_rule(
'tool_call',
format_literal("<tool_call>") + " space (" +
' | '.join(tool_rules) +
") space " + format_literal("</tool_call>"))# + ' space')
# Ideally we'd want a negative lookahead of /<tool\\?_call>/, but it's just too hard to express in GBNF for now.
# So we just over-constrain the content rule to not contain literals dangerously getting close to <tool_call>
content_rule = converter._add_rule('content', '[^<] | "<" [^t<] | "<t" [^o<]')
# content_rule = converter._add_rule('content', converter.not_literal('<tool_call>'))
converter._add_rule(
'root',
# tool_call_rule)
f'{content_rule}* ({tool_call_rule}+ {content_rule}*)?' if allow_parallel_calls \
else f'{content_rule}* {tool_call_rule}?')
self.grammar = converter.format_grammar()
# # Constrain the output to be a non-tool-call message (constrained to a JSON schema or not)
# # OR a tool-call message respecting the schema of any of the tools
# converter._add_rule(
# "root",
# converter._format_literal(prefix) + " (" +
# (response_rule or converter.not_literal("<tool_call>")) + " | " +
# converter._format_literal("<tool_call>") + " (" +
# ' | '.join(tool_rules) +
# ") " + converter._format_literal("</tool_call>") +
# ")") # + converter._format_literal(suffix))
@typechecked
def parse(self, s: str) -> Optional[Message]:
s = self.args.chat_template.strip_suffix(s)
if r'<tool\_call>' in s:
# Some weird escaping of underscores is happening w/ Mixtral 8x7B Instruct
s = s.replace(r'\_', '_')
parts = _tool_call_re.split(s)
if len(parts) == 1:
return Message(role="assistant", content=s)
else:
content = []
tool_calls = []
for i, part in enumerate(parts):
if i % 2 == 0:
content.append(part)
else:
try:
fc = json.loads(part)
except json.JSONDecodeError:
raise ValueError(f'Failed to parse tool call as JSON: {part}\nFull string: {s}')
tool_calls.append(
ToolCall(
id=gen_callid(),
function=FunctionCall(**fc)))
content = '\n'.join(content).strip()
return Message(role="assistant", content=content if content else None, tool_calls=tool_calls)
# if '<tool_call>'.startswith(ls) or ls.startswith('<tool_call>'):
# if ls.startswith('<tool_call>') and ls.endswith('</tool_call>' + suffix):
# tool_call = ls[len('<tool_call>'):-len('</tool_call>' + suffix)]
# return Message(role="assistant", content=None, tool_calls=[json.loads(tool_call)])
# return None
# else:
# return Message(role="assistant", content=s)
class TemplatedToolsChatHandler(ToolCallTagsChatHandler):
def __init__(self, args: ChatHandlerArgs, template: str, escapes_underscores=False, allow_parallel_calls=True):
super().__init__(args, escapes_underscores=escapes_underscores, allow_parallel_calls=allow_parallel_calls)
assert '{tools}' in template, 'Template must contain "{tools}"'
self.output_format_prompt = Message(
role="system",
content=template.replace(
'{tools}',
'\n'.join(json.dumps(tool.model_dump(), indent=2) for tool in self.args.tools),
)
)
class Hermes2ProToolsChatHandler(ToolCallTagsChatHandler):
def __init__(self, args: ChatHandlerArgs):
super().__init__(args, escapes_underscores=False, allow_parallel_calls=False)
# Hackily import https://github.com/NousResearch/Hermes-Function-Calling
path = str(Path(__file__).parent / "hermes_function_calling")
if path not in sys.path: sys.path.insert(0, path)
try:
from examples.openai.hermes_function_calling.prompter import PromptManager
except ImportError:
raise ImportError(f"Please `git clone https://github.com/NousResearch/Hermes-Function-Calling {path}`")
prompt = PromptManager().generate_prompt(user_prompt=[], tools=[json.dumps(tool) for tool in tools])
assert len(prompt) == 1 and prompt[0]["role"] == "system"
self.output_format_prompt = Message(**prompt[0])
class FunctionaryToolsChatHandler(ChatHandler):
def __init__(self, args: ChatHandlerArgs, allow_parallel_calls: bool):
super().__init__(args)
# Only allowing a single tool call at a time for now.
# Note that if there were more, they'd be separated by a '<|from|>assistant' literal
self.output_format_prompt = Message(
role="system",
content= '// Supported function definitions that should be called when necessary.\n' +
_tools_typescript_signatures(args.tools)
)
converter = SchemaConverter(prop_order={}, allow_fetch=False, dotall=False, raw_pattern=False)
tool_rules = [
converter._add_rule(
tool.function.name + '-call',
converter._format_literal(tool.function.name) + ' ' + converter._format_literal('\n<|content|>\n') + ' ' +
converter.visit(tool.function.parameters, tool.function.name + '-args') + ' ' +
converter._format_literal('\n'))
# converter.visit(
# dict(
# type="object",
# properties=dict(
# name=dict(const=tool.function.name),
# arguments=tool.function.parameters,
# ),
# required=['name', 'arguments']
# ),
# f'{tool.function.name}-tool-call'
# )
for i, tool in enumerate(self.args.tools)
]
not_from_rule = converter._add_rule('not_from', converter.not_literal("<|from|>"))
content_without_start_rule = converter._add_rule(
'content_without_start',
converter._format_literal("all\n<|content|>") + ' ' + not_from_rule + '*')
start_rule = converter._add_rule('start', converter._format_literal('<|from|>assistant\n<|recipient|>'))
content_rule = converter._add_rule('content', start_rule + ' ' + content_without_start_rule)
tool_call_without_start_rule = converter._add_rule(
'tool_call_without_start',
' | '.join(tool_rules))
# + ' ' +
# converter.not_literal("all", dotall=False) + ' ' + converter._format_literal('\n<|content|>\n') + ' ' + not_from_rule + '*')
tool_call_rule = converter._add_rule('tool_call', f'{start_rule} {tool_call_without_start_rule}')
# converter._add_rule('root', f'({content_without_start_rule} ({content_rule})* ({tool_call_rule}+ {content_rule}*)? | {tool_call_without_start_rule} (* {tool_call_rule}{content_rule}*')
converter._add_rule(
'root',
f'{content_without_start_rule} {content_rule}* ({tool_call_rule}+ {content_rule}*)? | '
f'{tool_call_without_start_rule} {tool_call_rule}* {content_rule}*' if allow_parallel_calls \
else f'{content_without_start_rule} {tool_call_rule}? | {tool_call_without_start_rule}')
self.grammar = converter.format_grammar()
# converter._add_rule(
# "root",
# converter._format_literal(prefix) + " (" +
# (response_rule or converter.not_literal("<|recipient|>")) + " | " +
# (' | '.join(
# converter._format_literal(f"<|recipient|>{tool.function.name}\n<|content|>") + " " +
# converter.visit(tool.function.parameters, tool.function.name + '-args')
# for tool in tools
# )) +
# ") " +
# ")") # + converter._format_literal(suffix))
@typechecked
def parse(self, s: str) -> Optional[Message]:
s = self.args.chat_template.strip_suffix(s)
parts = _recipient_content_re.split(s)
if len(parts) == 1:
return Message(role="assistant", content=s)
else:
text_content = []
tool_calls: list[ToolCall] = []
for i in range((len(parts) - 1) // 3):
assert parts[i * 3].strip() == '', f'Unexpected content before tool call: {parts[i * 3]}'
recipient = parts[i * 3 + 1].strip()
content = parts[i * 3 + 2]
if recipient == 'all':
text_content.append(content)
else:
try:
arguments = json.loads(content)
except json.JSONDecodeError:
raise ValueError(f'Failed to parse tool call content as JSON: {content}')
tool_calls.append(
ToolCall(
id=gen_callid(),
function=FunctionCall(name=recipient, arguments=arguments)))
assert parts[-1].strip() in ('', '<|stop|>'), f'Unexpected content after tool calls: {parts[-1]}\nFull string: {s}'
content = '\n'.join(text_content).strip()
return Message(role="assistant", content=content if content else None, tool_calls=tool_calls if tool_calls else None)
def _make_bespoke_schema(response_schema, tool_call_schema):
return {
"type": "object",
"properties": {
# "original_goal": {"title": "Original Goal", "type": "string"},
"thought": {
# "title": "Thought about how the next step brings us closer to achieving the original goal",
"type": "string"
},
"next_step": {
"title": "Next Step: either a result or one or more tool calls to achieve the original goal",
"oneOf": [
{
"title": "Tool Calls",
"properties": {
# "type": {
# "const": "tool_calls"
# },
"tool_calls": {
"type": "array",
"items": tool_call_schema
}
},
"required": ["tool_calls"]
},
{
"title": "Result (achieving original goal)",
"properties": {
"result": response_schema,
},
"required": ["result"]
},
]
},
},
"required": ["original_goal", "thought", "next_step"]
}
class BespokeToolsChatHandler(ChatHandler):
def __init__(self, args: ChatHandlerArgs):
super().__init__(args)
# args.response_schema = args.response_schema or {}
converter = SchemaConverter(prop_order={}, allow_fetch=False, dotall=False, raw_pattern=False)
response_schema = args.response_schema or {"type": "string"}
converter.visit(
_make_bespoke_schema(
response_schema,
{
"oneOf": [
{
"type": "object",
"properties": {
"name": {"const": tool.function.name},
"arguments": tool.function.parameters,
},
"required": ["name", "arguments"]
}
for tool in self.args.tools
]
}
),
'',
)
self.grammar = converter.format_grammar()
self.output_format_prompt = Message(
role="system",
content='\n'.join([
'You are a function calling AI model.',
'Here are the tools available:',
_tools_schema_signatures(self.args.tools, indent=2),
_please_respond_with_schema(
_make_bespoke_schema(
response_schema,
{
"properties": {
"name": {
"title": "Name of the tool to call",
"type": "string"
},
"arguments": {
"title": "Arguments to pass to the tool",
"type": "object"
}
},
"required": ["name", "arguments"]
}
)
),
])
)
@typechecked
def parse(self, s: str) -> Optional[Message]:
s = self.args.chat_template.strip_suffix(s)
try:
data = json.loads(s)
except json.JSONDecodeError:
raise ValueError(f'Failed to parse data as JSON: {s}')
next_step = data['next_step']
if 'result' in next_step:
return Message(role="assistant", content=json.dumps(next_step['result']))
elif 'tool_calls' in next_step:
return Message(
role="assistant",
content=data["thought"],
tool_calls=[
ToolCall(id=gen_callid(), function=FunctionCall(**tc))
for tc in next_step['tool_calls']
]
)
else:
raise ValueError(f'Unexpected data: {data}')
_SHORT_TEMPLATE='\n'.join([
'Here are the tools available:',
'<tools>',
'{tools}',
'</tools>',
])
_LONG_TEMPLATE='\n'.join([
# '''You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags.''',
'You may call one or more functions to assist with the user query. Don\'t make assumptions about what values to plug into functions. Here are the available tools:',
'<tools>',
'{tools}',
'</tools>',
'',
# 'Use the following json schema for each tool call you will make: {"properties": {"arguments": {"title": "Arguments", "type": "object"}, "name": {"title": "Name", "type": "string"}}, "required": ["arguments", "name"], "title": "FunctionCall", "type": "object"}',
# '',
# 'For each function call return a json object with function name and arguments within <tool_call></tool_call> XML tags as follows:',
'To call each function, give its name and arguments within <tool_call></tool_call> XML tags as follows:',
'<tool_call>',
'{"name": <function-name>, "arguments": <args-dict>}',
'</tool_call>',
# 'This is not hypothetical, you're not asked what you would do. If you need a tool called, just call it with <tool_call>...</tool_call>.''',
])
def get_chat_handler(args: ChatHandlerArgs, allow_parallel_calls=False) -> ChatHandler:
if not args.tools:
return NoToolsChatHandler(args)
elif args.chat_template.tool_style == ToolsPromptStyle.TYPESCRIPT_FUNCTIONARY_V2:
return FunctionaryToolsChatHandler(args)
elif args.chat_template.tool_style == ToolsPromptStyle.TOOLS_SHORT:
return TemplatedToolsChatHandler(args, _SHORT_TEMPLATE, allow_parallel_calls=allow_parallel_calls)
elif args.chat_template.tool_style == ToolsPromptStyle.TOOLS_LONG:
return TemplatedToolsChatHandler(args, _LONG_TEMPLATE, allow_parallel_calls=allow_parallel_calls)
elif args.chat_template.tool_style == ToolsPromptStyle.TOOLS_MISTRAL:
return TemplatedToolsChatHandler(args, _LONG_TEMPLATE, escapes_underscores=True, allow_parallel_calls=allow_parallel_calls)
elif args.chat_template.tool_style == ToolsPromptStyle.TOOLS_BESPOKE:
return BespokeToolsChatHandler(args)
elif args.chat_template.tool_style == ToolsPromptStyle.TOOLS_HERMES_2_PRO:
return Hermes2ProToolsChatHandler(args)
else:
raise ValueError(f"Unsupported tool call style: {args.chat_template.tool_style}")
_ts_converter = SchemaToTypeScriptConverter()
def _please_respond_with_schema(schema: dict) -> str:
# sig = json.dumps(schema, indent=2)
sig = _ts_converter.visit(schema)
return f'Please respond in JSON format with the following schema: {sig}'
def _tools_typescript_signatures(tools: list[Tool]) -> str:
return 'namespace functions {' + '\n'.join(
'// ' + tool.function.description.replace('\n', '\n// ') + '\n' + ''
'type ' + tool.function.name + ' = (_: ' + _ts_converter.visit(tool.function.parameters) + ") => any;\n"
for tool in tools
) + '} // namespace functions'
def _tools_schema_signatures(tools: list[Tool], indent=None) -> str:
return '\n'.join(
json.dumps(tool.model_dump(), indent=indent)
for tool in tools
)
_tool_call_re = re.compile(
'<tool_call>(.*?)</tool_call>', re.DOTALL)
_recipient_content_re = re.compile(r'(?:(?:<\|(?:stop|from)\|>)+ *assistant\n<\|recipient\|>|^) *([^ <|>\n]+) *\n<\|content\|>(.*?)(?:$|<\|stop\|>\s*$|(?=(?:<\|(?:stop|from)\|>)+ *assistant\n))', re.DOTALL)
def gen_callid():
return f'call_{random.randint(0, 1000000)}'