autoforge 1.8.1__tar.gz → 1.8.2__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (41) hide show
  1. {autoforge-1.8.1 → autoforge-1.8.2}/PKG-INFO +1 -1
  2. {autoforge-1.8.1 → autoforge-1.8.2}/pyproject.toml +1 -1
  3. {autoforge-1.8.1 → autoforge-1.8.2}/src/AutoForge.egg-info/PKG-INFO +1 -1
  4. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/OptimizerHelper.py +0 -35
  5. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/PruningHelper.py +3 -8
  6. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Modules/Optimizer.py +57 -17
  7. {autoforge-1.8.1 → autoforge-1.8.2}/README.md +0 -0
  8. {autoforge-1.8.1 → autoforge-1.8.2}/setup.cfg +0 -0
  9. {autoforge-1.8.1 → autoforge-1.8.2}/src/AutoForge.egg-info/SOURCES.txt +0 -0
  10. {autoforge-1.8.1 → autoforge-1.8.2}/src/AutoForge.egg-info/dependency_links.txt +0 -0
  11. {autoforge-1.8.1 → autoforge-1.8.2}/src/AutoForge.egg-info/entry_points.txt +0 -0
  12. {autoforge-1.8.1 → autoforge-1.8.2}/src/AutoForge.egg-info/requires.txt +0 -0
  13. {autoforge-1.8.1 → autoforge-1.8.2}/src/AutoForge.egg-info/top_level.txt +0 -0
  14. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Gui/__init__.py +0 -0
  15. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Gui/browser_gui/Views/__init__.py +0 -0
  16. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Gui/browser_gui/Views/backend_database.py +0 -0
  17. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Gui/browser_gui/Views/filament_editor.py +0 -0
  18. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Gui/browser_gui/app.py +0 -0
  19. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Gui/browser_gui/main.py +0 -0
  20. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/CAdamW.py +0 -0
  21. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/FilamentHelper.py +0 -0
  22. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/Heightmaps/ChristofidesHeightMap.py +0 -0
  23. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/Heightmaps/DepthEstimateHeightMap.py +0 -0
  24. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/Heightmaps/OldHeightmap.py +0 -0
  25. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/Heightmaps/__init__.py +0 -0
  26. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/ImageHelper.py +0 -0
  27. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/OutputHelper.py +0 -0
  28. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/__init__.py +0 -0
  29. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Helper/filamentcolors_library.py +0 -0
  30. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Init_comparer/__init__.py +0 -0
  31. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Init_comparer/auto_forge_init_check.py +0 -0
  32. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Init_comparer/plot_results.py +0 -0
  33. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Loss/LossFunctions.py +0 -0
  34. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Loss/PerceptionLoss.py +0 -0
  35. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Loss/__init__.py +0 -0
  36. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Misc/__init__.py +0 -0
  37. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Misc/color_fit_all.py +0 -0
  38. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/Modules/__init__.py +0 -0
  39. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/__init__.py +0 -0
  40. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/__main__.py +0 -0
  41. {autoforge-1.8.1 → autoforge-1.8.2}/src/autoforge/auto_forge.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: AutoForge
3
- Version: 1.8.1
3
+ Version: 1.8.2
4
4
  Summary: AutoForge is a Python tool for generating 3D printed layered models from an input image.
5
5
  Author-email: Hendric Voss <hvoss@techfak.uni-bielefeld.de>
6
6
  License: CC BY-NC-SA 4.0
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
4
4
 
5
5
  [project]
6
6
  name = "AutoForge"
7
- version="1.8.1"
7
+ version="1.8.2"
8
8
  description = "AutoForge is a Python tool for generating 3D printed layered models from an input image."
9
9
  authors = [{ name = "Hendric Voss", email = "hvoss@techfak.uni-bielefeld.de" }]
10
10
  readme = "README.md"
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: AutoForge
3
- Version: 1.8.1
3
+ Version: 1.8.2
4
4
  Summary: AutoForge is a Python tool for generating 3D printed layered models from an input image.
5
5
  Author-email: Hendric Voss <hvoss@techfak.uni-bielefeld.de>
6
6
  License: CC BY-NC-SA 4.0
@@ -277,38 +277,3 @@ def composite_image_disc(
277
277
  comp = comp + rem_after.unsqueeze(-1) * background # [H,W,3]
278
278
 
279
279
  return comp * 255.0
280
-
281
-
282
- def discretize_solution(
283
- params: dict, tau_global: float, h: float, max_layers: int, rng_seed: int = -1
284
- ):
285
- """
286
- Convert continuous logs to discrete layer counts and discrete color IDs.
287
-
288
- Args:
289
- params (dict): Dictionary containing the parameters 'pixel_height_logits' and 'global_logits'.
290
- tau_global (float): Temperature parameter for global material assignment.
291
- h (float): Height of each layer.
292
- max_layers (int): Maximum number of layers.
293
- rng_seed (int, optional): Random seed for deterministic sampling. Defaults to -1.
294
-
295
- Returns:
296
- tuple: A tuple containing:
297
- - torch.Tensor: Discrete global material assignments, shape [max_layers].
298
- - torch.Tensor: Discrete height image, shape [H, W].
299
- """
300
- pixel_height_logits = params["pixel_height_logits"]
301
- global_logits = params["global_logits"]
302
- pixel_heights = (max_layers * h) * torch.sigmoid(pixel_height_logits)
303
- discrete_height_image = torch.round(pixel_heights / h).to(torch.int32)
304
- discrete_height_image = torch.clamp(discrete_height_image, 0, max_layers)
305
-
306
- num_layers = global_logits.shape[0]
307
- discrete_global_vals = []
308
- for j in range(num_layers):
309
- p = deterministic_gumbel_softmax(
310
- global_logits[j], tau_global, hard=True, rng_seed=rng_seed + j
311
- )
312
- discrete_global_vals.append(torch.argmax(p))
313
- discrete_global = torch.stack(discrete_global_vals, dim=0)
314
- return discrete_global, discrete_height_image
@@ -6,7 +6,7 @@ from tqdm import tqdm
6
6
  from joblib import Parallel, delayed
7
7
  import threading
8
8
 
9
- from autoforge.Helper.OptimizerHelper import discretize_solution, composite_image_disc
9
+ from autoforge.Helper.OptimizerHelper import composite_image_disc
10
10
  from autoforge.Loss.LossFunctions import compute_loss
11
11
  from autoforge.Modules.Optimizer import FilamentOptimizer
12
12
 
@@ -349,16 +349,11 @@ def remove_layer_from_solution(
349
349
  new_max_layers (int): Updated number of layers.
350
350
  """
351
351
  # Get the current discrete height image (used to decide which pixels need adjusting)
352
-
353
352
  effective_logits = optimizer._apply_height_offset(
354
353
  params["pixel_height_logits"], params["height_offsets"]
355
354
  )
356
-
357
- _, disc_height = discretize_solution(
358
- {
359
- "pixel_height_logits": effective_logits,
360
- "global_logits": params["global_logits"],
361
- },
355
+ _, disc_height = optimizer.discretize_solution(
356
+ params,
362
357
  final_tau,
363
358
  h,
364
359
  current_max_layers,
@@ -9,9 +9,9 @@ from tqdm import tqdm
9
9
 
10
10
  from autoforge.Helper.CAdamW import CAdamW
11
11
  from autoforge.Helper.OptimizerHelper import (
12
- discretize_solution,
13
12
  composite_image_cont,
14
13
  composite_image_disc,
14
+ deterministic_gumbel_softmax,
15
15
  )
16
16
  from autoforge.Loss.LossFunctions import loss_fn, compute_loss
17
17
 
@@ -115,6 +115,7 @@ class FilamentOptimizer:
115
115
  self.params = {
116
116
  "pixel_height_logits": self.pixel_height_logits,
117
117
  "global_logits": global_logits_init,
118
+ "height_offsets": self.height_offsets,
118
119
  }
119
120
 
120
121
  # Tau schedule
@@ -303,6 +304,50 @@ class FilamentOptimizer:
303
304
 
304
305
  return loss
305
306
 
307
+ def discretize_solution(
308
+ self,
309
+ params: dict,
310
+ tau_global: float,
311
+ h: float,
312
+ max_layers: int,
313
+ rng_seed: int = -1,
314
+ ):
315
+ """
316
+ Convert continuous logs to discrete layer counts and discrete color IDs.
317
+
318
+ Args:
319
+ params (dict): Dictionary containing the parameters 'pixel_height_logits' and 'global_logits'.
320
+ tau_global (float): Temperature parameter for global material assignment.
321
+ h (float): Height of each layer.
322
+ max_layers (int): Maximum number of layers.
323
+ rng_seed (int, optional): Random seed for deterministic sampling. Defaults to -1.
324
+
325
+ Returns:
326
+ tuple: A tuple containing:
327
+ - torch.Tensor: Discrete global material assignments, shape [max_layers].
328
+ - torch.Tensor: Discrete height image, shape [H, W].
329
+ """
330
+ pixel_logits = params["pixel_height_logits"]
331
+ pixel_offset = params["height_offsets"]
332
+ effective_logits = self._apply_height_offset(
333
+ pixel_logits=pixel_logits, height_offsets=pixel_offset
334
+ )
335
+
336
+ global_logits = params["global_logits"]
337
+ pixel_heights = (max_layers * h) * torch.sigmoid(effective_logits)
338
+ discrete_height_image = torch.round(pixel_heights / h).to(torch.int32)
339
+ discrete_height_image = torch.clamp(discrete_height_image, 0, max_layers)
340
+
341
+ num_layers = global_logits.shape[0]
342
+ discrete_global_vals = []
343
+ for j in range(num_layers):
344
+ p = deterministic_gumbel_softmax(
345
+ global_logits[j], tau_global, hard=True, rng_seed=rng_seed + j
346
+ )
347
+ discrete_global_vals.append(torch.argmax(p))
348
+ discrete_global = torch.stack(discrete_global_vals, dim=0)
349
+ return discrete_global, discrete_height_image
350
+
306
351
  def log_to_tensorboard(
307
352
  self, interval: int = 100, namespace: str = "", step: int = None
308
353
  ):
@@ -489,13 +534,8 @@ class FilamentOptimizer:
489
534
  )
490
535
 
491
536
  if best:
492
- effective_logits = self._apply_height_offset(
493
- self.best_params["pixel_height_logits"],
494
- self.best_params["height_offsets"],
495
- )
496
- current_params["pixel_height_logits"] = effective_logits
497
- disc_global, disc_height_image = discretize_solution(
498
- current_params,
537
+ disc_global, disc_height_image = self.discretize_solution(
538
+ self.best_params,
499
539
  self.vis_tau,
500
540
  self.h,
501
541
  self.max_layers,
@@ -505,7 +545,7 @@ class FilamentOptimizer:
505
545
  else:
506
546
  tau_height, tau_global = self._get_tau()
507
547
  with torch.no_grad():
508
- disc_global, disc_height_image = discretize_solution(
548
+ disc_global, disc_height_image = self.discretize_solution(
509
549
  current_params,
510
550
  tau_height,
511
551
  self.h,
@@ -603,18 +643,14 @@ class FilamentOptimizer:
603
643
  tau_h, tau_g = self.vis_tau, self.vis_tau
604
644
  with torch.no_grad():
605
645
  effective_logits = self._apply_height_offset()
606
- params_with_offset = {
607
- "pixel_height_logits": effective_logits,
608
- "global_logits": self.params["global_logits"],
609
- }
610
- disc_global, disc_height_image = discretize_solution(
611
- params_with_offset, tau_g, self.h, self.max_layers, rng_seed=seed
646
+ disc_global, disc_height_image = self.discretize_solution(
647
+ self.params, tau_g, self.h, self.max_layers, rng_seed=seed
612
648
  )
613
649
 
614
650
  # 2) Compute discrete-mode composite
615
651
  with torch.no_grad():
616
652
  comp_disc = composite_image_disc(
617
- self.params["pixel_height_logits"],
653
+ effective_logits,
618
654
  self.params["global_logits"],
619
655
  self.vis_tau,
620
656
  self.vis_tau,
@@ -659,8 +695,12 @@ class FilamentOptimizer:
659
695
  best_loss = start_loss
660
696
  for i in tqdm(range(num_seeds), desc="Searching for new best seed"):
661
697
  seed = np.random.randint(0, 1000000)
662
- comp_disc = composite_image_disc(
698
+ effective_logits = self._apply_height_offset(
663
699
  self.best_params["pixel_height_logits"],
700
+ self.best_params["height_offsets"],
701
+ )
702
+ comp_disc = composite_image_disc(
703
+ effective_logits,
664
704
  self.best_params["global_logits"],
665
705
  self.vis_tau,
666
706
  self.vis_tau,
File without changes
File without changes