support sd3.5

examples/image_synthesis/sd35_text_to_image.py

@@ -0,0 +1,19 @@
+from diffsynth import ModelManager, SD3ImagePipeline
+import torch
+
+
+
+model_manager = ModelManager(torch_dtype=torch.bfloat16, device="cuda", model_id_list=["StableDiffusion3.5-large"])
+pipe = SD3ImagePipeline.from_model_manager(model_manager)
+
+prompt = "A capybara holding a sign that reads Hello World"
+negative_prompt = ""
+
+image = pipe(
+    prompt=prompt,
+    negative_prompt=negative_prompt,
+    cfg_scale=3.5,
+    num_inference_steps=28, width=1024, height=1024,
+    seed=0
+)
+image.save("image_1024.jpg")