new images

run.py

@@ -5,6 +5,8 @@ import PIL.Image
 import numpy as np
 import tqdm
 
+seed = 3
+
 # 1. Unroll the full loop
 # ==================================================================
 # load all models
@@ -19,7 +21,7 @@ unet.to(torch_device)
 vqvae.to(torch_device)
 
 # generate gaussian noise to be decoded
-generator = torch.manual_seed(0)
+generator = torch.manual_seed(seed)
 noise = torch.randn(
     (1, unet.in_channels, unet.image_size, unet.image_size),
     generator=generator,
@@ -50,8 +52,6 @@ image_processed = (image_processed + 1.0) * 127.5
 image_processed = image_processed.numpy().astype(np.uint8)
 image_pil = PIL.Image.fromarray(image_processed[0])
 
-image_pil.save("generated_image_unrolled.png")
-
 
 # 2. Use pipeline
 # ==================================================================
@@ -64,7 +64,7 @@ import tqdm
 pipeline = LatentDiffusionUncondPipeline.from_pretrained("./")
 
 # generatae image by calling the pipeline
-generator = torch.manual_seed(0)
+generator = torch.manual_seed(seed)
 image = pipeline(generator=generator, num_inference_steps=200)["sample"]
 
 # process image
@@ -73,4 +73,4 @@ image_processed = (image_processed + 1.0) * 127.5
 image_processed = image_processed.numpy().astype(np.uint8)
 image_pil = PIL.Image.fromarray(image_processed[0])
 
-image_pil.save("generated_image_pipeline.png")
+image_pil.save(f"generated_image_{seed}.png")