Skip to content

fix: llama_types & unit tests for functionary #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: functionary-v2.5
Choose a base branch
from
Open

fix: llama_types & unit tests for functionary #1

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions llama_cpp/llama_types.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -222,6 +222,7 @@ class ChatCompletionRequestToolMessage(TypedDict):
role: Literal["tool"]
content: Optional[str]
tool_call_id: str
name: str


class ChatCompletionRequestFunctionMessage(TypedDict):
Expand Down
259 changes: 259 additions & 0 deletions tests/test_functionary.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,259 @@
from typing import cast

import llama_cpp.llama_types as llama_types
from llama_cpp import llama

from llama_cpp.llama_chat_format import FunctionaryV25ChatHandler, functionary_new_chat_handler
from unittest.mock import Mock

from llama_cpp.llama_tokenizer import LlamaHFTokenizer


class LlamaMock:
tokenizer_ = LlamaHFTokenizer.from_pretrained('meetkai/functionary-small-v2.5-GGUF')
create_completion = Mock()
verbose = True


def test_functionary_v25_generate_text_only():
llama_mock = LlamaMock()

llama_mock.create_completion.return_value = llama_types.CreateCompletionResponse(
id='cmpl-0963a3b2-e0d5-4dac-948f-74fdc6875140',
object='text_completion',
created=1723029767,
model='functionary-small-v2.5.Q4_0.gguf',
choices=[
llama_types.CompletionChoice(
text='The capital of France is Paris.',
index=0,
logprobs=None,
finish_reason='stop',
),
],
usage=llama_types.CompletionUsage(
prompt_tokens=82,
completion_tokens=7,
total_tokens=89,
),
)

response: llama_types.CreateChatCompletionResponse = functionary_new_chat_handler(
llama=cast(llama.Llama, llama_mock),
messages=[
llama_types.ChatCompletionRequestUserMessage(
role='user',
content='what is the capital of France'
),
],
stream=False
)

# Ensure the llama_mock has been called
llama_mock.create_completion.assert_called_once()

assert len(response.get('choices')) == 1

message: llama_types.ChatCompletionResponseMessage = response.get('choices')[0].get('message')
assert message.get('content') == 'The capital of France is Paris.'
assert message.get('role') == 'assistant'


def test_functionary_v25_generate_tools():
llama_mock = LlamaMock()

llama_mock.create_completion.side_effect = [
# Should use tool or text
llama_types.CreateCompletionResponse(
id='cmpl-f25ed8fd-85bd-49e6-b3a6-cc827774764f',
object='text_completion',
created=1723032052,
model='/models/functionary-small-v2.5.Q4_0.gguf',
choices=[
llama_types.CompletionChoice(
text=FunctionaryV25ChatHandler.tool_call_token,
index=0,
logprobs=None,
finish_reason='length',
),
],
usage=llama_types.CompletionUsage(
prompt_tokens=121,
completion_tokens=1,
total_tokens=122,
),
),
# which tool to uses
llama_types.CreateCompletionResponse(
id='cmpl-397f36e4-06ff-4f3e-8a00-f5069d2f3cac',
object='text_completion',
created=1723032055,
model='/models/functionary-small-v2.5.Q4_0.gguf',
choices=[
llama_types.CompletionChoice(
text='get_current_weather',
index=0,
logprobs=None,
finish_reason='stop',
),
],
usage=llama_types.CompletionUsage(
prompt_tokens=122,
completion_tokens=4,
total_tokens=126,
),
),
# determine arguments
llama_types.CreateCompletionResponse(
id='cmpl-702420e2-ceac-4658-9872-941b2a7f88dd',
object='text_completion',
created=1723032055,
model='/models/functionary-small-v2.5.Q4_0.gguf',
choices=[
llama_types.CompletionChoice(
text='{"location": "Hanoi, Vietnam"}',
index=0,
logprobs=None,
finish_reason='stop',
),
],
usage=llama_types.CompletionUsage(
prompt_tokens=126,
completion_tokens=9,
total_tokens=135,
),
),
# should continue
llama_types.CreateCompletionResponse(
id= 'cmpl-d17805ca-8f84-466c-8440-8c04587ed5d2',
object= 'text_completion',
created=1723032057,
model= '/models/functionary-small-v2.5.Q4_0.gguf',
choices=[
llama_types.CompletionChoice(
text='',
index=0,
logprobs=None,
finish_reason='stop',
),
],
usage=llama_types.CompletionUsage(
prompt_tokens=135,
completion_tokens=0,
total_tokens=135,
),
),
]

response: llama_types.CreateChatCompletionResponse = functionary_new_chat_handler(
llama=cast(llama.Llama, llama_mock),
messages=[
llama_types.ChatCompletionRequestUserMessage(
role='user',
content="what's the weather like in Hanoi?",
),
],
stream=False,
tools=[
llama_types.ChatCompletionTool(
type='function',
function=llama_types.ChatCompletionToolFunction(
name='get_current_weather',
description='Get the current weather',
parameters={
'type': 'object',
'properties': {
'location': {
'type': 'string',
'description': 'The city and state, e.g., San Francisco, CA'
},
},
'required': ['location']
}
)
)
]
)

# Ensure the llama_mock has been called
llama_mock.create_completion.assert_called()

assert len(response.get('choices')) == 1

# should be calling tools
assert response.get('choices')[0].get('finish_reason') == 'tool_calls'

message: llama_types.ChatCompletionResponseMessage = response.get('choices')[0].get('message')

assert message.get('role') == 'assistant'

tool_calls: llama_types.ChatCompletionMessageToolCalls = message.get('tool_calls')
assert len(tool_calls) == 1

assert tool_calls[0].get('type') == 'function'
assert tool_calls[0].get('function').get('name') == 'get_current_weather'
assert tool_calls[0].get('function').get('arguments') == '{"location": "Hanoi, Vietnam"}'


def test_functionary_v25_assemble_messages():
llama_mock = LlamaMock()

llama_mock.create_completion.return_value = llama_types.CreateCompletionResponse(
id='cmpl-6ac62660-6010-4eee-aa82-0d763f7101de',
object='text_completion',
created=1723035815,
model='functionary-small-v2.5.Q4_0.gguf',
choices=[
llama_types.CompletionChoice(
text='The current weather in Hanoi, Vietnam is 37°C.',
index=0,
logprobs=None,
finish_reason='stop',
),
],
usage=llama_types.CompletionUsage(
prompt_tokens=112,
completion_tokens=18,
total_tokens=130,
),
)

response: llama_types.CreateChatCompletionResponse = functionary_new_chat_handler(
llama=cast(llama.Llama, llama_mock),
messages=[
llama_types.ChatCompletionRequestUserMessage(
role='user',
content="what's the weather like in Hanoi?",
),
llama_types.ChatCompletionRequestAssistantMessage(
role='assistant',
content=None,
tool_calls=[
llama_types.ChatCompletionMessageToolCall(
id='call_EPXRajVAJ0TlvzeZ80v4uiYn',
type='function',
function=llama_types.ChatCompletionMessageToolCallFunction(
name='get_current_weather',
arguments='{"location": "Hanoi, Vietnam"}'
),
),
],
),
llama_types.ChatCompletionRequestToolMessage(
role='tool',
content='{"location": "Hanoi, Vietnam", "temperature": "unknown"}',
name='get_current_weather',
tool_call_id='call_EPXRajVAJ0TlvzeZ80v4uiYn',
),
],
stream=False,
)

# Ensure the llama_mock has been called
llama_mock.create_completion.assert_called_once()

assert len(response.get('choices')) == 1

message: llama_types.ChatCompletionResponseMessage = response.get('choices')[0].get('message')
assert message.get('content') == 'The current weather in Hanoi, Vietnam is 37°C.'
assert message.get('role') == 'assistant'