File size: 8,400 Bytes
9c31777 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 |
# coding=utf-8
# Copyright 2024 HuggingFace Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
import PIL.Image
import pytest
from smolagents import (
CodeAgent,
RunResult,
ToolCallingAgent,
stream_to_gradio,
)
from smolagents.models import (
ChatMessage,
ChatMessageToolCall,
ChatMessageToolCallFunction,
MessageRole,
Model,
TokenUsage,
)
class FakeLLMModel(Model):
def __init__(self, give_token_usage: bool = True):
self.give_token_usage = give_token_usage
def generate(self, prompt, tools_to_call_from=None, **kwargs):
if tools_to_call_from is not None:
return ChatMessage(
role=MessageRole.ASSISTANT,
content="",
tool_calls=[
ChatMessageToolCall(
id="fake_id",
type="function",
function=ChatMessageToolCallFunction(name="final_answer", arguments={"answer": "image"}),
)
],
token_usage=TokenUsage(input_tokens=10, output_tokens=20) if self.give_token_usage else None,
)
else:
return ChatMessage(
role=MessageRole.ASSISTANT,
content="""<code>
final_answer('This is the final answer.')
</code>""",
token_usage=TokenUsage(input_tokens=10, output_tokens=20) if self.give_token_usage else None,
)
class MonitoringTester(unittest.TestCase):
def test_code_agent_metrics(self):
agent = CodeAgent(
tools=[],
model=FakeLLMModel(),
max_steps=1,
)
agent.run("Fake task")
self.assertEqual(agent.monitor.total_input_token_count, 10)
self.assertEqual(agent.monitor.total_output_token_count, 20)
def test_toolcalling_agent_metrics(self):
agent = ToolCallingAgent(
tools=[],
model=FakeLLMModel(),
max_steps=1,
)
agent.run("Fake task")
self.assertEqual(agent.monitor.total_input_token_count, 10)
self.assertEqual(agent.monitor.total_output_token_count, 20)
def test_code_agent_metrics_max_steps(self):
class FakeLLMModelMalformedAnswer(Model):
def generate(self, prompt, **kwargs):
return ChatMessage(
role=MessageRole.ASSISTANT,
content="Malformed answer",
token_usage=TokenUsage(input_tokens=10, output_tokens=20),
)
agent = CodeAgent(
tools=[],
model=FakeLLMModelMalformedAnswer(),
max_steps=1,
)
agent.run("Fake task")
self.assertEqual(agent.monitor.total_input_token_count, 20)
self.assertEqual(agent.monitor.total_output_token_count, 40)
def test_code_agent_metrics_generation_error(self):
class FakeLLMModelGenerationException(Model):
def generate(self, prompt, **kwargs):
raise Exception("Cannot generate")
agent = CodeAgent(
tools=[],
model=FakeLLMModelGenerationException(),
max_steps=1,
)
with pytest.raises(Exception) as e:
agent.run("Fake task")
assert "Cannot generate" in str(e.value)
def test_streaming_agent_text_output(self):
agent = CodeAgent(
tools=[],
model=FakeLLMModel(),
max_steps=1,
planning_interval=2,
)
# Use stream_to_gradio to capture the output
outputs = list(stream_to_gradio(agent, task="Test task"))
self.assertEqual(len(outputs), 11)
plan_message = outputs[1]
self.assertEqual(plan_message.role, "assistant")
self.assertIn("<code>", plan_message.content)
final_message = outputs[-1]
self.assertEqual(final_message.role, "assistant")
self.assertIn("This is the final answer.", final_message.content)
def test_streaming_agent_image_output(self):
agent = ToolCallingAgent(
tools=[],
model=FakeLLMModel(),
max_steps=1,
verbosity_level=100,
)
# Use stream_to_gradio to capture the output
outputs = list(
stream_to_gradio(
agent,
task="Test task",
additional_args=dict(image=PIL.Image.new("RGB", (100, 100))),
)
)
self.assertEqual(len(outputs), 7)
final_message = outputs[-1]
self.assertEqual(final_message.role, "assistant")
self.assertIsInstance(final_message.content, dict)
self.assertEqual(final_message.content["mime_type"], "image/png")
def test_streaming_with_agent_error(self):
class DummyModel(Model):
def generate(self, prompt, **kwargs):
return ChatMessage(role=MessageRole.ASSISTANT, content="Malformed call")
agent = CodeAgent(
tools=[],
model=DummyModel(),
max_steps=1,
)
# Use stream_to_gradio to capture the output
outputs = list(stream_to_gradio(agent, task="Test task"))
self.assertEqual(len(outputs), 11)
final_message = outputs[-1]
self.assertEqual(final_message.role, "assistant")
self.assertIn("Malformed call", final_message.content)
def test_run_return_full_result(self):
agent = CodeAgent(
tools=[],
model=FakeLLMModel(),
max_steps=1,
return_full_result=True,
)
result = agent.run("Fake task")
self.assertIsInstance(result, RunResult)
self.assertEqual(result.output, "This is the final answer.")
self.assertEqual(result.state, "success")
self.assertEqual(result.token_usage, TokenUsage(input_tokens=10, output_tokens=20))
self.assertIsInstance(result.messages, list)
self.assertGreater(result.timing.duration, 0)
agent = ToolCallingAgent(
tools=[],
model=FakeLLMModel(),
max_steps=1,
return_full_result=True,
)
result = agent.run("Fake task")
self.assertIsInstance(result, RunResult)
self.assertEqual(result.output, "image")
self.assertEqual(result.state, "success")
self.assertEqual(result.token_usage, TokenUsage(input_tokens=10, output_tokens=20))
self.assertIsInstance(result.messages, list)
self.assertGreater(result.timing.duration, 0)
# Below 2 lines should be removed when the attributes are removed
assert agent.monitor.total_input_token_count == 10
assert agent.monitor.total_output_token_count == 20
def test_run_result_no_token_usage(self):
agent = CodeAgent(
tools=[],
model=FakeLLMModel(give_token_usage=False),
max_steps=1,
return_full_result=True,
)
result = agent.run("Fake task")
self.assertIsInstance(result, RunResult)
self.assertEqual(result.output, "This is the final answer.")
self.assertEqual(result.state, "success")
self.assertIsNone(result.token_usage)
self.assertIsInstance(result.messages, list)
self.assertGreater(result.timing.duration, 0)
agent = ToolCallingAgent(
tools=[],
model=FakeLLMModel(give_token_usage=False),
max_steps=1,
return_full_result=True,
)
result = agent.run("Fake task")
self.assertIsInstance(result, RunResult)
self.assertEqual(result.output, "image")
self.assertEqual(result.state, "success")
self.assertIsNone(result.token_usage)
self.assertIsInstance(result.messages, list)
self.assertGreater(result.timing.duration, 0)
|