File size: 14,741 Bytes
bc50ea0 | 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 | #!/usr/bin/env python3
"""Chapter geometry and existing-lock compatibility, using temporary projects."""
import copy
import importlib
import json
import math
import pathlib
import runpy
import tempfile
import unittest
from unittest.mock import patch
import torch
fixture = runpy.run_path(str(pathlib.Path(__file__).with_name(
"_chapter_delivery_unit_test.py")))
chain = fixture["chain"]
folders = fixture["folder_paths"]
class ChapterResolutionTest(unittest.TestCase):
def setUp(self):
self.temp = tempfile.TemporaryDirectory()
self.addCleanup(self.temp.cleanup)
folders.output_directory = self.temp.name
self.root = pathlib.Path(self.temp.name)
self.raw = {
"shots": [{"id": f"scene_{i:02d}", "prompt": f"Scene {i}",
"length": 39, "context_length": 0}
for i in range(1, 7)],
"chapters": fixture["make_editorial"]()["chapters"],
}
def plan(self, width=64, height=64, raw=None, mode="guide"):
return chain._normalize_plan(
json.dumps(self.raw if raw is None else raw), "chapter_delivery",
width, height, 5, "video", "head", "disabled", "generated_audio",
5, 1.0, 8, 11, 18, "test-model", 0, mode)
def locks(self, indices, sizes=None):
old = self.plan()
editorial = fixture["make_editorial"]()
editorial["locked_scene_ids"] = [f"scene_{i:02d}" for i in indices]
for i in indices:
metadata = {
"segment": {"id": f"scene_{i:02d}", "index": i},
"compatibility": dict(old["compatibility"]),
"scene_dependency": chain._scene_dependency_record(old, i, None),
}
if sizes and i in sizes:
metadata["scene_dependency"]["scopes"]["global_generation"].update(sizes[i])
chain._atomic_json(chain._artifact_paths(old, i)["metadata"], metadata)
chain._atomic_json(chain._run_editorial_path("chapter_delivery"), editorial)
return old
def test_inherit_preserves_legacy_hashes(self):
plan = self.plan()
plain = copy.deepcopy(self.raw)
plain.pop("chapters")
self.assertEqual(plan["plan_hash"], self.plan(raw=plain)["plan_hash"])
self.assertTrue(all("resolution" not in shot for shot in plan["shots"]))
self.raw["chapters"][0]["title"] = "Changed notes"
self.assertEqual(plan["plan_hash"], self.plan()["plan_hash"])
def test_chapter_dimensions_and_dependency_isolation(self):
old = self.plan()
self.raw["chapters"][1]["resolution"] = {"width": 128, "height": 96}
new = self.plan()
self.assertEqual(chain.scene_resolution(new, 1), {"width": 64, "height": 64})
self.assertEqual(chain.scene_resolution(new, 4), {"width": 128, "height": 96})
for index in range(1, 7):
diffs = chain._scene_dependency_diffs(
chain._scene_dependency_record(old, index, None),
chain._scene_dependency_record(new, index, None))
self.assertEqual(bool(diffs), index >= 4)
result = chain.MiniMaxH3ChainCurrent().current({"plan": new, "index": 4})
self.assertEqual(result["result"][8:10], (128, 96))
self.assertEqual(new["compatibility"]["width"], 64)
def test_inherit_is_plan_default_not_previous_chapter(self):
self.raw["chapters"][0]["resolution"] = {"width": 128, "height": 96}
self.assertEqual(chain.scene_resolution(self.plan(), 4), {"width": 64, "height": 64})
def test_old_locks_pin_only_their_chapter_without_writes(self):
old = self.locks([1, 2, 3])
before = {path: path.read_bytes() for path in self.root.rglob("*.json")}
new = self.plan(128, 96)
for index in range(1, 4):
self.assertEqual(chain.scene_resolution(new, index), {"width": 64, "height": 64})
self.assertEqual(chain._scene_dependency_diffs(
chain._scene_dependency_record(old, index, None),
chain._scene_dependency_record(new, index, None)), [])
self.assertEqual(chain.scene_resolution(new, 4), {"width": 128, "height": 96})
self.assertEqual(before, {path: path.read_bytes() for path in self.root.rglob("*.json")})
recovered = self.plan(128, 96, raw=chain._effective_editor_plan(new))
self.assertEqual([chain.scene_resolution(recovered, i) for i in range(1, 7)],
[chain.scene_resolution(new, i) for i in range(1, 7)])
def test_explicit_conflict_rejected_and_unlock_allows_change(self):
self.locks([1, 2, 3])
self.raw["chapters"][0]["resolution"] = {"width": 128, "height": 96}
with self.assertRaisesRegex(ValueError, "pinned.*64x64.*Unlock"):
self.plan()
chain._atomic_json(chain._run_editorial_path("chapter_delivery"), fixture["make_editorial"]())
self.assertEqual(chain.scene_resolution(self.plan(), 1), {"width": 128, "height": 96})
def test_conflicting_saved_sizes_in_one_chapter_rejected(self):
self.locks([1, 2], {2: {"width": 128, "height": 96}})
with self.assertRaisesRegex(ValueError, "locked saved scenes at different resolutions"):
self.plan()
def test_lock_does_not_waive_prompt_or_model_identity(self):
old = self.locks([1])
self.raw["shots"][0]["prompt"] = "Changed identity"
new = self.plan(128, 96)
self.assertTrue(chain._scene_dependency_diffs(
chain._scene_dependency_record(old, 1, None),
chain._scene_dependency_record(new, 1, None)))
new["compatibility"]["generation_fingerprint"] = "other-model"
self.assertTrue(chain._scene_dependency_diffs(
chain._scene_dependency_record(old, 2, None),
chain._scene_dependency_record(new, 2, None)))
def test_pin_before_first_chapter_survives_recovery(self):
self.locks([1])
self.raw["chapters"] = self.raw["chapters"][1:]
plan = self.plan(128, 96)
restored_raw = chain._effective_editor_plan(plan)
self.assertEqual(restored_raw["chapters"][0]["resolution"], {"width": 64, "height": 64})
chain._atomic_json(chain._run_editorial_path("chapter_delivery"), fixture["make_editorial"]())
self.assertEqual(chain.scene_resolution(self.plan(128, 96, raw=restored_raw), 1),
{"width": 64, "height": 64})
def test_native_context_cross_size_rejected_in_preflight(self):
self.raw["chapters"][1]["resolution"] = {"width": 128, "height": 96}
self.raw["shots"][3]["context_length"] = 5
self.raw["shots"][3]["audio_context_length"] = 0
for mode in ("masked_av", "audio_feathered_av", "latent_guide"):
with self.subTest(mode=mode):
plan = self.plan(mode=mode)
with self.assertRaisesRegex(ValueError, "across different chapter resolutions"):
chain._validate_scene_resolution_boundary(plan, 4)
_, report = chain._preflight_chain(plan, scene_range="4", verify_resume_history=False)
self.assertFalse(report["ok"])
self.assertIn("across different chapter resolutions", json.dumps(report))
self.raw["shots"][3]["context_length"] = 0
chain._validate_scene_resolution_boundary(self.plan(mode="masked_av"), 4)
def test_new_lock_fences_an_already_queued_plan(self):
queued = self.plan(128, 96)
self.locks([1])
with self.assertRaisesRegex(ValueError, "differs from the queued Plan"):
chain._require_chapter_resolution_locks(queued)
# Resolving again picks up the saved Chapter 1 size without changing Chapter 2.
chain._require_chapter_resolution_locks(self.plan(128, 96))
def test_missing_geometry_and_stale_identity_are_not_silently_adopted(self):
plan = self.locks([1])
path = chain._artifact_paths(plan, 1)["metadata"]
metadata = chain._read_json(path)
metadata.pop("scene_dependency")
metadata.pop("compatibility")
chain._atomic_json(path, metadata)
with self.assertRaisesRegex(ValueError, "Cannot verify the saved resolution"):
self.plan()
metadata["segment"]["id"] = "wrong_scene"
chain._atomic_json(path, metadata)
with self.assertRaisesRegex(ValueError, "no longer matches"):
self.plan()
def test_context_rejects_wrong_conditioning_geometry(self):
self.raw["chapters"][1]["resolution"] = {"width": 128, "height": 96}
plan = self.plan()
with patch.object(chain, "_streams_from_latent", return_value=(torch.zeros(1, 4, 2, 4, 4), None)):
with self.assertRaisesRegex(ValueError, "Wire Current Shot width/height"):
chain.MiniMaxH3ChainContext().apply({"plan": plan, "index": 4}, [], None, {})
with self.assertRaisesRegex(ValueError, "decoded dimensions"):
chain.MiniMaxH3ChainSegmentSave().save(
{"plan": plan, "index": 4}, torch.zeros(39, 64, 64, 3), {})
def test_chapter_exports_retain_each_saved_resolution(self):
segments = [fixture["make_segment"](self.root, i) for i in range(1, 7)]
for i, segment in enumerate(segments):
# Existing checkpoints expose geometry through their dependency record.
segment["scene_dependency"] = {"scopes": {"global_generation": {
"width": 64 if i < 3 else 128, "height": 64 if i < 3 else 96}}}
manifest = fixture["make_manifest"](segments, True)
manifest["compatibility"].update(width=128, height=96)
chain._atomic_json(chain._run_editorial_path("chapter_delivery"), fixture["make_editorial"]())
one, path = chain._chapter_manifest_from_manifest(manifest, 1)
before = pathlib.Path(path).read_bytes()
two, _ = chain._chapter_manifest_from_manifest(manifest, 2)
self.assertEqual(one["compatibility"]["width"], 64)
self.assertEqual(two["compatibility"]["width"], 128)
self.assertEqual(before, pathlib.Path(path).read_bytes())
with self.assertRaisesRegex(ValueError, "Export each chapter separately"):
chain.MiniMaxH3ChainAssemble().assemble(manifest, "none", "mixed", 192)
def test_real_checkpoints_resume_chapter_exports_and_review_preview(self):
self.raw["shots"] = [
{"id": f"scene_{i:02d}", "prompt": f"Scene {i}", "length": 5,
"context_length": 0, "audio_context_length": 0} for i in (1, 2)]
self.raw["chapters"][1]["start_scene_id"] = "scene_02"
self.raw["chapters"][1]["resolution"] = {"width": 128, "height": 96}
plan = self.plan()
editorial = {"scene_order": [{"scene": i, "scene_id": f"scene_{i:02d}"} for i in (1, 2)],
"chapters": self.raw["chapters"]}
chain._atomic_json(chain._run_editorial_path("chapter_delivery"), editorial)
with patch.object(chain, "_streams_from_latent", side_effect=lambda latent: latent["samples"]):
for index in (1, 2):
state = chain._initial_state(plan, index)
width, height = (chain.scene_resolution(plan, index)[key] for key in ("width", "height"))
latent = {"samples": [torch.zeros(1, 24, 2, height // 16, width // 16),
torch.zeros(1, 32, 2, 9)]}
audio = {"waveform": torch.zeros(1, 2, round(5 / 24 * 8000)), "sample_rate": 8000}
segment = chain.MiniMaxH3ChainSegmentSave().save(
state, torch.full((5, height, width, 3), index / 3), latent, audio)["result"][0]
if index == 1:
original = pathlib.Path(chain._artifact_paths(plan, 1)["metadata"]).read_bytes()
else:
self.assertEqual(segment["resolution"], {"width": 128, "height": 96})
self.assertEqual(original, pathlib.Path(chain._artifact_paths(plan, 1)["metadata"]).read_bytes())
manifest = chain._partial_manifest(state, segment)
for number in (1, 2):
chapter, _ = chain._chapter_manifest_from_manifest(manifest, number)
result = chain.MiniMaxH3ChainAssemble().assemble(chapter, "none", "chapter-test", 192)
with chain.av.open(result["result"][0]) as container:
frames = list(container.decode(video=0))
self.assertEqual(len(frames), 5)
self.assertEqual((frames[0].width, frames[0].height), (64, 64) if number == 1 else (128, 96))
preview, warning = chain._assemble_review_partial(state, segment, "none", None)
self.assertFalse(warning)
with chain.av.open(preview) as container:
frames = list(container.decode(video=0))
self.assertEqual(len(frames), 5, "mixed-size review displays the current chapter only")
self.assertEqual((frames[0].width, frames[0].height), (128, 96))
self.assertEqual(len(state["segments"]), 1, "review does not truncate execution history")
def test_deferred_upscale_reads_source_scene_geometry(self):
upscale = importlib.import_module(chain.__package__ + ".upscale_nodes")
source = {"index": 1, "raw_frames": 5, "delivered_frames": 5,
"resolution": {"width": 64, "height": 96}}
state = {"index": 1, "source_manifest": {
"segments": [source], "compatibility": {"width": 128, "height": 128}}}
streams = [torch.zeros(1, 24, 2, 6, 4), torch.zeros(1, 32, 2, 9)]
with patch.object(upscale, "_load_source_tensors", return_value={}), \
patch.object(upscale, "_source_latent", return_value=({"samples": streams}, "test")), \
patch.object(chain, "_streams_from_latent", return_value=streams):
result = upscale.MiniMaxH3ChainUpscaleCurrent().current(state)
self.assertEqual(result[7:9], (64, 96))
def test_validation_and_cache_invalidation(self):
for value in ({"width": 65, "height": 64}, {"width": True, "height": 64},
{"width": 64}, {"width": 0, "height": 64}, "128x96"):
self.raw["chapters"][0]["resolution"] = value
with self.subTest(value=value), self.assertRaises(ValueError):
self.plan()
self.raw["chapters"][0].pop("resolution")
serialized = json.dumps(self.raw)
self.assertTrue(math.isnan(chain.MiniMaxH3ChainPlan.IS_CHANGED(serialized)))
self.assertTrue(math.isnan(chain.MiniMaxH3ChainPlanStudio.IS_CHANGED(plan_json=serialized)))
if __name__ == "__main__":
unittest.main()
|