Spaces:
Running on Zero
Running on Zero
File size: 19,769 Bytes
b701455 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 | """
Shared pytest fixtures and mock utilities for LightDiffusion test suite.
This module provides:
- Mock checkpoint loaders that don't load real model weights
- Mock model patchers and CLIP models
- Common utilities for testing without GPU dependencies
"""
import os
import sys
from pathlib import Path
from unittest.mock import MagicMock, patch, PropertyMock
from typing import Tuple, Dict, Any, Optional, List
import pytest
import pytest_asyncio
import torch
import numpy as np
# Add project root to path
project_root = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(project_root))
# =============================================================================
# Mock Model Classes
# =============================================================================
class MockModelPatcher:
"""Mock model patcher that simulates model loading without actual weights.
This provides the same interface as the real ModelPatcher but doesn't
load any actual model data, making tests fast and memory-efficient.
"""
def __init__(self, model_name: str = "mock_model", model_type: str = "SD15"):
self.model_name = model_name
self.model_type = model_type
self.model = MagicMock()
self.model.diffusion_model = MagicMock()
self.model.model_options = {}
self.model.model_type = 0 # EPS
self.model.model_sampling = MagicMock()
self.model.model_sampling.sigma_min = 0.02
self.model.model_sampling.sigma_max = 14.6
self.model.model_sampling.sigmas = torch.linspace(0.02, 14.6, 1000)
# Provide a simple sigma function for tests that accepts tensor inputs
def _sigma(t):
# Ensure tensor input and return tensor of same shape filled with mean sigma
try:
t_t = torch.as_tensor(t)
mean_sigma = float(self.model.model_sampling.sigmas.mean())
return torch.full_like(t_t, mean_sigma, dtype=torch.float32)
except Exception:
return float(self.model.model_sampling.sigmas.mean())
self.model.model_sampling.sigma = _sigma
self.model.model_sampling.timestep = lambda x: x * 1000
# Ensure the inner mock model provides memory sizing helpers that match
# what production model objects expose. This prevents MagicMock values
# from leaking into memory calculations during tests.
self.model.memory_required = lambda shape: 1024 * 1024 * 1024 # 1GB default for tests
self.model.model_memory_required = lambda device=None: 2 * 1024 * 1024 * 1024 # 2GB
# Provide a simple apply_model implementation that returns a real tensor
# with the same shape as the input to avoid propagation of MagicMock values
# into conditioning and sampling logic.
def _apply_model(input_x, timestep, **kwargs):
return torch.randn_like(input_x)
self.model.apply_model = _apply_model
self.latent_format = MagicMock()
self.latent_format.latent_channels = 4
self.patches = {}
self.object_patches = {}
self.weight_inplace_update = False
self.load_device = torch.device("cpu")
self.offload_device = torch.device("cpu")
self.current_device = torch.device("cpu")
self.model_options = {}
# Mirror important model attributes expected by Device and ModelPatcher
self.model.model_loaded_weight_memory = 0
self.model.model_lowvram = False
def model_dtype(self):
return torch.float16
def memory_required(self, shape):
return 1024 * 1024 * 1024 # 1GB
def model_memory_required(self, device=None):
return 2 * 1024 * 1024 * 1024 # 2GB
# ------------------------------------------------------------------
# Methods to emulate ModelPatcher behavior (used by Device and pipeline)
# ------------------------------------------------------------------
def model_size(self) -> int:
"""Return the mocked total model size in bytes.
Default to 2GB to simulate a moderate-sized model for memory
calculations in tests.
"""
return 2 * 1024 * 1024 * 1024 # 2GB
def loaded_size(self) -> int:
"""Return the size of currently loaded weights.
Defaults to the tracked attribute on the inner mock model.
"""
return getattr(self.model, "model_loaded_weight_memory", 0)
def model_patches_to(self, device):
"""No-op in the mock; present for interface compatibility."""
self.current_device = device
def patch_model(self, device_to=None, patch_weights=True):
"""Return the inner mock model to simulate patching behavior."""
return self.model
def unpatch_model(self, device_to=None, unpatch_weights=True):
"""No-op unpatch in the mock."""
return
def partially_load(self, device_to, extra_memory=0):
"""Simulate partially loading model weights into memory.
Increments the recorded loaded weight memory by up to extra_memory
but never exceeding the model's total mocked size.
"""
prev = getattr(self.model, "model_loaded_weight_memory", 0)
add = min(extra_memory, max(0, self.model_size() - prev))
self.model.model_loaded_weight_memory = prev + add
return self.model.model_loaded_weight_memory - prev
def get_model_object(self, name):
if name == "model_sampling":
return self.model.model_sampling
return MagicMock()
def clone(self):
"""Return a clone of this patcher."""
cloned = MockModelPatcher(self.model_name, self.model_type)
cloned.patches = self.patches.copy()
cloned.object_patches = self.object_patches.copy()
return cloned
def add_patches(self, patches: Dict, strength: float = 1.0):
"""Add patches (LoRA weights, etc.)."""
self.patches.update(patches)
def get_model_object(self, name: str):
"""Get a model object by name.
This mirrors the behavior of the real ModelPatcher.get_model_object and
returns reasonable objects for names commonly used in tests.
"""
if name == "model_sampling":
return self.model.model_sampling
if name == "latent_format":
return self.latent_format
# Fall back to attributes on the inner mock model
return getattr(self.model, name, MagicMock())
def set_model_option(self, key: str, value: Any):
"""Set a model option."""
self.model_options[key] = value
def apply_tome(self, ratio: float = 0.5, max_downsample: int = 1) -> bool:
"""Mock ToMe application."""
return True
def remove_tome(self):
"""Mock ToMe removal."""
pass
def apply_stable_fast(self, enable_cuda_graph: bool = True):
"""Mock StableFast application."""
self.model.apply_stable_fast()
return self
def apply_deepcache(self, interval, depth, start, end):
"""Mock DeepCache application."""
self.model.apply_deepcache()
return self
class MockCLIP:
"""Mock CLIP model for testing text encoding without loading real weights."""
def __init__(self, clip_type: str = "SD15"):
self.clip_type = clip_type
self.cond_stage_model = MagicMock()
self.tokenizer = MagicMock()
self.layer_idx = -2
self.patcher = MagicMock()
def encode(self, text: str) -> Tuple[torch.Tensor, Dict]:
"""Mock encode that returns fake embeddings."""
# Return fake conditioning tensor (batch, seq_len, embed_dim)
if self.clip_type == "SDXL":
embed_dim = 2048 # SDXL uses concatenated L+G (768+1280)
else:
embed_dim = 768 # SD1.5
cond = torch.randn(1, 77, embed_dim)
pooled = torch.randn(1, embed_dim) if self.clip_type == "SDXL" else None
return cond, {"pooled_output": pooled}
def tokenize(self, text: str) -> Dict:
"""Mock tokenize."""
return {"input_ids": torch.randint(0, 49407, (1, 77))}
def encode_token_weights(self, tokens: Any) -> Tuple:
"""Mock encode_token_weights."""
if self.clip_type == "SDXL":
embed_dim = 2048
else:
embed_dim = 768
cond = torch.randn(1, 77, embed_dim)
pooled = torch.randn(1, embed_dim) if self.clip_type == "SDXL" else None
return cond, pooled
def encode_from_tokens(self, tokens: Dict, return_pooled: bool = False):
"""Encode directly from tokenized inputs.
This mirrors the interface used by CLIPTextEncode and the pipeline.
For tests we return random tensors with the expected shapes.
"""
if self.clip_type == "SDXL":
embed_dim = 2048
else:
embed_dim = 768
cond = torch.randn(1, 77, embed_dim)
pooled = torch.randn(1, embed_dim) if self.clip_type == "SDXL" else None
return (cond, pooled) if return_pooled else cond
def clone(self):
"""Clone the CLIP model, preserving layer index and other state."""
cloned = MockCLIP(self.clip_type)
cloned.layer_idx = self.layer_idx
return cloned
def clip_layer(self, stop_at_clip_layer: int):
"""Set the CLIP layer used for skip/prompt settings (no-op for mocks)."""
# The real CLIP implementation changes internal behavior when internal
# layers are skipped. For testing we simply record the configured
# layer index so that code using this API can inspect it if needed.
self.layer_idx = stop_at_clip_layer
return None
class MockVAE:
"""Mock VAE for testing encode/decode without real model weights."""
def __init__(self, latent_channels: int = 4):
self.latent_channels = latent_channels
self.first_stage_model = MagicMock()
self.latent_channels = latent_channels
def encode(self, images: torch.Tensor, flux: bool = False, **kwargs) -> torch.Tensor:
"""Encode images to latent space.
Accepts the same signature as the real VAE encode method (including
optional 'flux' flag) and returns a tensor of shape
[B, latent_channels, H/8, W/8].
"""
# Convert shape to expected format in case caller passes CPU tensors
batch = images.shape[0]
height = images.shape[1]
width = images.shape[2]
latent_h = height // 8
latent_w = width // 8
return torch.randn(batch, self.latent_channels, latent_h, latent_w)
def decode(self, latents: torch.Tensor, **kwargs) -> torch.Tensor:
"""Decode latents to images.
Accepts extra kwargs for compatibility with different VAE implementations.
"""
batch, channels, latent_h, latent_w = latents.shape
height = latent_h * 8
width = latent_w * 8
return torch.randn(batch, 3, height, width)
class MockCheckpointResult:
"""Container for mock checkpoint loading results."""
def __init__(self, model_type: str = "SD15"):
self.model_patcher = MockModelPatcher("mock_checkpoint", model_type)
self.clip = MockCLIP(model_type)
self.vae = MockVAE()
def as_tuple(self) -> Tuple:
"""Return as tuple matching CheckpointLoaderSimple output."""
return (self.model_patcher, self.clip, self.vae)
# =============================================================================
# Mock Loader Classes
# =============================================================================
class MockCheckpointLoaderSimple:
"""Mock checkpoint loader that doesn't load real model files.
Use this when you want to test code that calls CheckpointLoaderSimple
without actually loading 6GB model files.
"""
def __init__(self):
self.loaded_checkpoints = []
def load_checkpoint(
self,
ckpt_name: str,
output_vae: bool = True,
output_clip: bool = True
) -> Tuple:
"""Load a mock checkpoint.
Args:
ckpt_name: Path/name of checkpoint (used to detect model type)
output_vae: Whether to return VAE
output_clip: Whether to return CLIP
Returns:
Tuple of (model_patcher, clip, vae)
"""
self.loaded_checkpoints.append(ckpt_name)
# Detect model type from filename
ckpt_lower = ckpt_name.lower()
if "sdxl" in ckpt_lower or "xl" in ckpt_lower:
model_type = "SDXL"
elif "flux" in ckpt_lower:
model_type = "FLUX"
else:
model_type = "SD15"
result = MockCheckpointResult(model_type)
return result.as_tuple()
class MockUnetLoaderGGUF:
"""Mock GGUF UNet loader for Flux models."""
def __init__(self):
self.loaded_models = []
def load_unet(
self,
unet_name: str,
dequant_dtype: Optional[str] = None,
patch_dtype: Optional[str] = None
) -> Tuple:
"""Load a mock GGUF UNet."""
self.loaded_models.append(unet_name)
return (MockModelPatcher(unet_name, "FLUX"),)
# =============================================================================
# Pytest Fixtures
# =============================================================================
@pytest.fixture
def mock_checkpoint_loader():
"""Provide a MockCheckpointLoaderSimple instance."""
return MockCheckpointLoaderSimple()
@pytest.fixture
def mock_model_patcher():
"""Provide a MockModelPatcher instance."""
return MockModelPatcher()
@pytest.fixture
def mock_sd15_checkpoint():
"""Provide mock SD1.5 checkpoint result."""
return MockCheckpointResult("SD15")
@pytest.fixture
def mock_sdxl_checkpoint():
"""Provide mock SDXL checkpoint result."""
return MockCheckpointResult("SDXL")
@pytest.fixture
def mock_flux_checkpoint():
"""Provide mock Flux checkpoint result."""
return MockCheckpointResult("FLUX")
@pytest.fixture
def mock_clip_sd15():
"""Provide mock SD1.5 CLIP model."""
return MockCLIP("SD15")
@pytest.fixture
def server_client():
"""FastAPI TestClient for in-process server endpoint testing.
Use this fixture in API/integration tests to avoid starting a subprocess.
"""
from fastapi.testclient import TestClient
import server as _server
return TestClient(_server.app)
@pytest_asyncio.fixture
async def async_server_client():
"""Async in-process HTTP client for FastAPI endpoint testing.
Uses HTTPX's ASGI transport because the bundled TestClient currently
deadlocks against this app in the sandboxed test environment.
"""
import httpx
import server as _server
await _server.startup_event()
await _server._start_buffer()
transport = httpx.ASGITransport(app=_server.app)
async with httpx.AsyncClient(
transport=transport,
base_url="http://testserver",
) as client:
yield client
worker_task = getattr(_server._generation_buffer, "_worker_task", None)
if worker_task is not None:
worker_task.cancel()
try:
await worker_task
except BaseException:
pass
@pytest.fixture
def mock_clip_sdxl():
"""Provide mock SDXL CLIP model."""
return MockCLIP("SDXL")
@pytest.fixture
def mock_vae():
"""Provide mock VAE model."""
return MockVAE()
@pytest.fixture
def sample_latent_4ch():
"""Provide sample 4-channel latent tensor (SD1.5/SDXL)."""
return {"samples": torch.randn(1, 4, 64, 64)}
@pytest.fixture
def sample_latent_16ch():
"""Provide sample 16-channel latent tensor (Flux)."""
return {"samples": torch.randn(1, 16, 64, 64)}
@pytest.fixture
def sample_image_tensor():
"""Provide sample image tensor (B, H, W, C) normalized 0-1."""
return torch.rand(1, 512, 512, 3)
@pytest.fixture
def patch_checkpoint_loader():
"""Context manager to patch CheckpointLoaderSimple globally."""
with patch(
"src.FileManaging.Loader.CheckpointLoaderSimple",
MockCheckpointLoaderSimple
) as mock:
yield mock
@pytest.fixture
def patch_model_loader():
"""Patch load_model_for_pipeline to return mock results."""
def mock_load(model_path=None, flux_dequant_dtype=None, flux_patch_dtype=None):
if model_path and "flux" in model_path.lower():
return ("FLUX", (MockModelPatcher(model_path, "FLUX"),))
elif model_path and "sdxl" in model_path.lower():
return ("SDXL", MockCheckpointResult("SDXL").as_tuple())
else:
return ("SD15", MockCheckpointResult("SD15").as_tuple())
with patch(
"src.user.model_loader.load_model_for_pipeline",
side_effect=mock_load
) as mock:
yield mock
@pytest.fixture
def temp_model_path(tmp_path):
"""Create a temporary mock model file path."""
model_file = tmp_path / "test_model.safetensors"
model_file.touch() # Create empty file
return str(model_file)
# =============================================================================
# Utility Functions
# =============================================================================
def create_mock_conditioning(
batch_size: int = 1,
seq_len: int = 77,
embed_dim: int = 768,
model_type: str = "SD15"
) -> List:
"""Create mock conditioning entries matching pipeline format.
Args:
batch_size: Number of conditions
seq_len: Sequence length
embed_dim: Embedding dimension
model_type: "SD15" (768), "SDXL" (2048), or "FLUX"
Returns:
List of [tensor, metadata_dict] entries
"""
if model_type == "SDXL":
embed_dim = 2048
elif model_type == "FLUX":
embed_dim = 4096
entries = []
for i in range(batch_size):
cond_tensor = torch.randn(1, seq_len, embed_dim)
meta = {"batch_index": [i]}
if model_type == "SDXL":
meta["pooled_output"] = torch.randn(1, 1280)
entries.append([cond_tensor, meta])
return entries
def assert_tensor_shape(tensor: torch.Tensor, expected_shape: Tuple):
"""Assert tensor has expected shape with informative error message."""
assert tensor.shape == expected_shape, (
f"Expected shape {expected_shape}, got {tensor.shape}"
)
def assert_valid_latent(latent_dict: Dict, expected_channels: int = 4):
"""Assert latent dictionary is valid."""
assert "samples" in latent_dict, "Latent dict must have 'samples' key"
samples = latent_dict["samples"]
assert samples.ndim == 4, f"Latent must be 4D, got {samples.ndim}D"
assert samples.shape[1] == expected_channels, (
f"Expected {expected_channels} channels, got {samples.shape[1]}"
)
# =============================================================================
# Global Hooks
# =============================================================================
def pytest_runtest_teardown(item, nextitem):
"""Ensure all patches are stopped after each test."""
patch.stopall()
def get_test_data_path(relative_path: str) -> Path:
"""Get absolute path to test data file."""
return project_root / "tests" / "data" / relative_path
def get_checkpoint_path(model_name: str) -> str:
"""Get path to checkpoint (returns mock path for testing)."""
return str(project_root / "include" / "checkpoints" / model_name)
# Create test data directory if needed
(project_root / "tests" / "data").mkdir(parents=True, exist_ok=True)
|