deprecate LatentDiffusionAutoencoder's decode_latents

Author: Laurent

docs/guides/adapting_sdxl/index.md

@@ -92,7 +92,7 @@ with no_grad():  # Disable gradient calculation for memory-efficient inference
             pooled_text_embedding=pooled_text_embedding,
             time_ids=time_ids,
         )
-    predicted_image = sdxl.lda.decode_latents(x)
+    predicted_image = sdxl.lda.latents_to_image(x)
 
 predicted_image.save("vanilla_sdxl.png")
 
@@ -145,7 +145,7 @@ predicted_image.save("vanilla_sdxl.png")
                 pooled_text_embedding=pooled_text_embedding,
                 time_ids=time_ids,
             )
-        predicted_image = sdxl.lda.decode_latents(x)
+        predicted_image = sdxl.lda.latents_to_image(x)
 
     predicted_image.save("vanilla_sdxl.png")
 
@@ -318,7 +318,7 @@ manager.add_loras("pixel-art-lora", load_from_safetensors("pixel-art-xl-v1.1.saf
                 pooled_text_embedding=pooled_text_embedding,
                 time_ids=time_ids,
             )
-        predicted_image = sdxl.lda.decode_latents(x)
+        predicted_image = sdxl.lda.latents_to_image(x)
 
     predicted_image.save("scifi_pixel_sdxl.png")
 
@@ -453,7 +453,7 @@ with torch.no_grad():
                 pooled_text_embedding=pooled_text_embedding,
                 time_ids=time_ids,
             )
-        predicted_image = sdxl.lda.decode_latents(x)
+        predicted_image = sdxl.lda.latents_to_image(x)
 
     predicted_image.save("scifi_pixel_IP_sdxl.png")
 
@@ -591,7 +591,7 @@ with torch.no_grad():
                 pooled_text_embedding=pooled_text_embedding,
                 time_ids=time_ids,
             )
-        predicted_image = sdxl.lda.decode_latents(x)
+        predicted_image = sdxl.lda.latents_to_image(x)
 
     predicted_image.save("scifi_pixel_IP_T2I_sdxl.png")
 

src/refiners/foundationals/latent_diffusion/auto_encoder.py

@@ -369,11 +369,6 @@ def images_to_latents(self, images: list[Image.Image]) -> Tensor:
         x = 2 * x - 1
         return self.encode(x)
 
-    # backward-compatibility alias
-    # TODO: deprecate this method
-    def decode_latents(self, x: Tensor) -> Image.Image:
-        return self.latents_to_image(x)
-
     def latents_to_image(self, x: Tensor) -> Image.Image:
         """
         Decode latents to an image.

src/refiners/foundationals/latent_diffusion/stable_diffusion_1/model.py

@@ -61,7 +61,7 @@ class StableDiffusion_1(LatentDiffusionModel):
         for step in sd15.steps:
             x = sd15(x, step=step, clip_text_embedding=clip_text_embedding)
 
-        predicted_image = sd15.lda.decode_latents(x)
+        predicted_image = sd15.lda.latents_to_image(x)
         predicted_image.save("output.png")
     ```
     """

tests/e2e/test_diffusion.py

@@ -1496,7 +1496,7 @@ def test_diffusion_sdxl_control_lora(
         )
 
     # decode latent to image
-    predicted_image = sdxl.lda.decode_latents(x)
+    predicted_image = sdxl.lda.latents_to_image(x)
 
     # ensure the predicted image is similar to the expected image
     ensure_similar_images(
@@ -1935,7 +1935,7 @@ def test_diffusion_ip_adapter_multi(
             clip_text_embedding=clip_text_embedding,
             condition_scale=7.5,
         )
-    predicted_image = sd15.lda.decode_latents(x)
+    predicted_image = sd15.lda.latents_to_image(x)
 
     ensure_similar_images(predicted_image, expected_image_ip_adapter_multi, min_psnr=43, min_ssim=0.98)
 
@@ -2245,7 +2245,7 @@ def test_diffusion_sdxl_sliced_attention(
             condition_scale=5,
         )
 
-    predicted_image = sdxl.lda.decode_latents(x)
+    predicted_image = sdxl.lda.latents_to_image(x)
     ensure_similar_images(predicted_image, expected_image, min_psnr=35, min_ssim=0.98)
 
 
@@ -2279,7 +2279,7 @@ def test_diffusion_sdxl_euler_deterministic(
             condition_scale=5,
         )
 
-    predicted_image = sdxl.lda.decode_latents(x)
+    predicted_image = sdxl.lda.latents_to_image(x)
     ensure_similar_images(predicted_image, expected_image)
 
 
@@ -2604,7 +2604,7 @@ def test_style_aligned(
         )
 
     # decode latents
-    predicted_images = [sdxl.lda.decode_latents(latent.unsqueeze(0)) for latent in x]
+    predicted_images = sdxl.lda.latents_to_images(x)
 
     # tile all images horizontally
     merged_image = Image.new("RGB", (1024 * len(predicted_images), 1024))

tests/e2e/test_doc_examples.py

@@ -109,7 +109,7 @@ def test_guide_adapting_sdxl_vanilla(
             time_ids=time_ids,
         )
 
-    predicted_image = sdxl.lda.decode_latents(x)
+    predicted_image = sdxl.lda.latents_to_image(x)
     ensure_similar_images(predicted_image, expected_image, min_psnr=35, min_ssim=0.98)
 
 
@@ -151,7 +151,7 @@ def test_guide_adapting_sdxl_single_lora(
             time_ids=time_ids,
         )
 
-    predicted_image = sdxl.lda.decode_latents(x)
+    predicted_image = sdxl.lda.latents_to_image(x)
     ensure_similar_images(predicted_image, expected_image, min_psnr=38, min_ssim=0.98)
 
 
@@ -195,7 +195,7 @@ def test_guide_adapting_sdxl_multiple_loras(
             time_ids=time_ids,
         )
 
-    predicted_image = sdxl.lda.decode_latents(x)
+    predicted_image = sdxl.lda.latents_to_image(x)
     ensure_similar_images(predicted_image, expected_image, min_psnr=38, min_ssim=0.98)
 
 
@@ -255,7 +255,7 @@ def test_guide_adapting_sdxl_loras_ip_adapter(
             time_ids=time_ids,
         )
 
-    predicted_image = sdxl.lda.decode_latents(x)
+    predicted_image = sdxl.lda.latents_to_image(x)
     ensure_similar_images(predicted_image, expected_image, min_psnr=29, min_ssim=0.98)