Spaces:
Sleeping
Sleeping
File size: 13,745 Bytes
cb922bf | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 | """
Property-based tests for AsyncEnvWrapper.
This module validates that AsyncEnvWrapper correctly wraps CloudResourceEnv
with async methods while preserving all return values and behavior.
Feature: final-submission-compliance
Properties tested:
- Property 1: AsyncEnvWrapper Reset Equivalence
"""
import pytest
import asyncio
import numpy as np
from hypothesis import given, strategies as st, settings
from env.environment import CloudResourceEnv
from env.async_wrapper import AsyncEnvWrapper
# Feature: final-submission-compliance, Property 1: AsyncEnvWrapper Reset Equivalence
@given(
task_name=st.sampled_from(["easy", "medium", "hard"]),
seed=st.integers(min_value=0, max_value=2**31 - 1)
)
@settings(max_examples=100, deadline=None)
def test_async_reset_equivalence(task_name, seed):
"""
Property 1: For any CloudResourceEnv instance, calling async reset() through
AsyncEnvWrapper should return the same observation as calling synchronous
reset() directly on the environment.
Validates: Requirements 2.2
"""
# Create two identical environments with same seed
env1 = CloudResourceEnv(task_name=task_name)
env1.rng = np.random.RandomState(seed)
env2 = CloudResourceEnv(task_name=task_name)
env2.rng = np.random.RandomState(seed)
# Wrap second environment with AsyncEnvWrapper
async_env = AsyncEnvWrapper(env2)
# Get synchronous reset observation
sync_obs = env1.reset()
# Get async reset observation
async_obs = asyncio.run(async_env.reset())
# Verify observations are identical
np.testing.assert_array_equal(
sync_obs,
async_obs,
err_msg=f"Async reset observation differs from sync reset for task={task_name}, seed={seed}"
)
# Verify observation shape and dtype
assert sync_obs.shape == async_obs.shape, "Observation shapes differ"
assert sync_obs.dtype == async_obs.dtype, "Observation dtypes differ"
# Verify observation has expected structure [cpu, memory, request_rate, resources]
assert len(async_obs) == 4, "Observation should have 4 elements"
# Verify all values are finite
assert np.all(np.isfinite(async_obs)), "Observation contains non-finite values"
@pytest.mark.parametrize("task_name", ["easy", "medium", "hard"])
def test_async_reset_preserves_state_structure(task_name):
"""
Verify that async reset returns observation with correct structure.
This test ensures AsyncEnvWrapper.reset() returns a numpy array with
the expected shape, dtype, and value ranges.
"""
env = CloudResourceEnv(task_name=task_name)
async_env = AsyncEnvWrapper(env)
obs = asyncio.run(async_env.reset())
# Verify type and shape
assert isinstance(obs, np.ndarray), "Observation should be numpy array"
assert obs.shape == (4,), "Observation should have shape (4,)"
assert obs.dtype == np.float32, "Observation should be float32"
# Verify value ranges
cpu_util, memory_util, request_rate, allocated_resources = obs
assert 0.0 <= cpu_util <= 100.0, f"CPU utilization {cpu_util} out of range [0, 100]"
assert 0.0 <= memory_util <= 100.0, f"Memory utilization {memory_util} out of range [0, 100]"
assert request_rate >= 0.0, f"Request rate {request_rate} should be non-negative"
assert allocated_resources >= 1.0, f"Allocated resources {allocated_resources} should be >= 1"
def test_async_reset_multiple_calls():
"""
Verify that multiple async reset calls produce different observations
due to stochastic initialization.
This test ensures AsyncEnvWrapper correctly propagates the stochastic
behavior of CloudResourceEnv.reset().
"""
env = CloudResourceEnv(task_name="medium")
async_env = AsyncEnvWrapper(env)
observations = []
for _ in range(10):
obs = asyncio.run(async_env.reset())
observations.append(obs.copy())
# Verify at least some observations differ (stochastic initialization)
# We check request_rate (index 2) which has stochastic variation
request_rates = [obs[2] for obs in observations]
unique_rates = set(request_rates)
# With 10 resets, we expect some variation in request rates
assert len(unique_rates) > 1, "Expected stochastic variation in reset observations"
def test_async_reset_after_episode():
"""
Verify that async reset works correctly after an episode completes.
This test ensures AsyncEnvWrapper.reset() can be called after step()
to start a new episode.
"""
env = CloudResourceEnv(task_name="medium")
async_env = AsyncEnvWrapper(env)
# Run initial episode
obs1 = asyncio.run(async_env.reset())
# Take a few steps
for _ in range(5):
obs, reward, done, info = asyncio.run(async_env.step(1)) # maintain action
if done:
break
# Reset for new episode
obs2 = asyncio.run(async_env.reset())
# Verify reset returns valid observation
assert isinstance(obs2, np.ndarray), "Reset should return numpy array"
assert obs2.shape == (4,), "Reset observation should have shape (4,)"
assert np.all(np.isfinite(obs2)), "Reset observation should contain finite values"
# Feature: final-submission-compliance, Property 2: AsyncEnvWrapper Step Equivalence
@given(
task_name=st.sampled_from(["easy", "medium", "hard"]),
seed=st.integers(min_value=0, max_value=2**31 - 1),
action=st.integers(min_value=0, max_value=2)
)
@settings(max_examples=100, deadline=None)
def test_async_step_equivalence(task_name, seed, action):
"""
Property 2: For any CloudResourceEnv instance and any valid action, calling
async step(action) through AsyncEnvWrapper should return the same
(observation, reward, done, info) tuple as calling synchronous step(action)
directly on the environment.
Validates: Requirements 2.3
"""
# Create two identical environments with same seed
env1 = CloudResourceEnv(task_name=task_name)
env1.rng = np.random.RandomState(seed)
env2 = CloudResourceEnv(task_name=task_name)
env2.rng = np.random.RandomState(seed)
# Reset both environments to same initial state
env1.reset()
env2.reset()
# Wrap second environment with AsyncEnvWrapper
async_env = AsyncEnvWrapper(env2)
# Execute synchronous step
sync_obs, sync_reward, sync_done, sync_info = env1.step(action)
# Execute async step
async_obs, async_reward, async_done, async_info = asyncio.run(async_env.step(action))
# Verify observations are identical
np.testing.assert_array_equal(
sync_obs,
async_obs,
err_msg=f"Async step observation differs from sync step for task={task_name}, seed={seed}, action={action}"
)
# Verify rewards are identical
assert sync_reward == async_reward, \
f"Async step reward {async_reward} differs from sync step reward {sync_reward}"
# Verify done flags are identical
assert sync_done == async_done, \
f"Async step done {async_done} differs from sync step done {sync_done}"
# Verify info dictionaries have same keys
assert set(sync_info.keys()) == set(async_info.keys()), \
f"Info dictionary keys differ: sync={set(sync_info.keys())}, async={set(async_info.keys())}"
# Verify info dictionary values are identical
for key in sync_info.keys():
if isinstance(sync_info[key], float):
# Use approximate equality for floating point values
assert abs(sync_info[key] - async_info[key]) < 1e-6, \
f"Info[{key}] differs: sync={sync_info[key]}, async={async_info[key]}"
else:
assert sync_info[key] == async_info[key], \
f"Info[{key}] differs: sync={sync_info[key]}, async={async_info[key]}"
@pytest.mark.parametrize("task_name", ["easy", "medium", "hard"])
@pytest.mark.parametrize("action", [0, 1, 2])
def test_async_step_preserves_tuple_structure(task_name, action):
"""
Verify that async step returns tuple with correct structure.
This test ensures AsyncEnvWrapper.step() returns a 4-tuple with
(observation, reward, done, info) matching expected types.
"""
env = CloudResourceEnv(task_name=task_name)
async_env = AsyncEnvWrapper(env)
asyncio.run(async_env.reset())
obs, reward, done, info = asyncio.run(async_env.step(action))
# Verify observation type and shape
assert isinstance(obs, np.ndarray), "Observation should be numpy array"
assert obs.shape == (4,), "Observation should have shape (4,)"
assert obs.dtype == np.float32, "Observation should be float32"
# Verify reward type
assert isinstance(reward, (int, float)), "Reward should be numeric"
assert np.isfinite(reward), "Reward should be finite"
# Verify done type
assert isinstance(done, bool), "Done should be boolean"
# Verify info type and structure
assert isinstance(info, dict), "Info should be dictionary"
expected_keys = {'step', 'cumulative_reward', 'cpu_util', 'memory_util',
'request_rate', 'allocated_resources', 'latency'}
assert set(info.keys()) == expected_keys, \
f"Info keys {set(info.keys())} differ from expected {expected_keys}"
def test_async_step_sequence_equivalence():
"""
Verify that a sequence of async steps produces same results as sync steps.
This test ensures AsyncEnvWrapper maintains equivalence across multiple
steps in an episode, not just a single step.
"""
seed = 42
task_name = "medium"
actions = [1, 2, 1, 0, 1] # sequence of actions
# Create two identical environments
env1 = CloudResourceEnv(task_name=task_name)
env1.rng = np.random.RandomState(seed)
env2 = CloudResourceEnv(task_name=task_name)
env2.rng = np.random.RandomState(seed)
async_env = AsyncEnvWrapper(env2)
# Reset both environments
sync_obs = env1.reset()
async_obs = asyncio.run(async_env.reset())
np.testing.assert_array_equal(sync_obs, async_obs,
err_msg="Initial observations differ")
# Execute sequence of steps
for i, action in enumerate(actions):
sync_obs, sync_reward, sync_done, sync_info = env1.step(action)
async_obs, async_reward, async_done, async_info = asyncio.run(async_env.step(action))
# Verify equivalence at each step
np.testing.assert_array_equal(
sync_obs, async_obs,
err_msg=f"Observations differ at step {i+1}"
)
assert sync_reward == async_reward, \
f"Rewards differ at step {i+1}: sync={sync_reward}, async={async_reward}"
assert sync_done == async_done, \
f"Done flags differ at step {i+1}: sync={sync_done}, async={async_done}"
# If episode terminated, stop
if sync_done:
break
def test_async_step_invalid_action():
"""
Verify that async step raises ValueError for invalid actions.
This test ensures AsyncEnvWrapper correctly propagates exceptions
from the wrapped environment.
"""
env = CloudResourceEnv(task_name="medium")
async_env = AsyncEnvWrapper(env)
asyncio.run(async_env.reset())
# Test invalid action values
invalid_actions = [-1, 3, 10, -5]
for invalid_action in invalid_actions:
with pytest.raises(ValueError, match="Invalid action"):
asyncio.run(async_env.step(invalid_action))
def test_async_step_after_done():
"""
Verify that async step raises RuntimeError when called after episode termination.
This test ensures AsyncEnvWrapper correctly propagates the done state
and prevents steps after termination.
"""
env = CloudResourceEnv(task_name="medium")
async_env = AsyncEnvWrapper(env)
asyncio.run(async_env.reset())
# Run until episode terminates
max_steps = 100
for _ in range(max_steps):
obs, reward, done, info = asyncio.run(async_env.step(1))
if done:
break
# Verify we reached done state
assert done, "Episode should have terminated within max_steps"
# Attempt to step after done should raise RuntimeError
with pytest.raises(RuntimeError, match="Cannot call step.*after episode termination"):
asyncio.run(async_env.step(1))
def test_async_step_preserves_episode_history():
"""
Verify that async step correctly updates episode history in wrapped environment.
This test ensures AsyncEnvWrapper doesn't interfere with the environment's
internal state tracking.
"""
env = CloudResourceEnv(task_name="medium")
async_env = AsyncEnvWrapper(env)
asyncio.run(async_env.reset())
# Execute several steps
actions = [1, 2, 0, 1]
for action in actions:
asyncio.run(async_env.step(action))
# Verify episode history was updated
assert len(env.episode_actions) == len(actions), \
f"Expected {len(actions)} actions in history, got {len(env.episode_actions)}"
assert env.episode_actions == actions, \
f"Episode actions {env.episode_actions} differ from expected {actions}"
assert len(env.episode_rewards) == len(actions), \
f"Expected {len(actions)} rewards in history, got {len(env.episode_rewards)}"
# Environment stores states after each step (not including initial state from reset)
assert len(env.episode_states) == len(actions), \
f"Expected {len(actions)} states in history, got {len(env.episode_states)}"
|