test local text-to-image models and cloudflare API

3d-generation-pipeline/generate_image_local.py

@@ -0,0 +1,28 @@
+import torch
+from diffusers import StableDiffusionPipeline, StableDiffusion3Pipeline
+import time
+
+start_timestamp = time.time()
+#model = "stabilityai/stable-diffusion-3.5-medium" # generation time: 13 min
+model = "stabilityai/stable-diffusion-3-medium-diffusers" # generation time: 10 min
+#model = "stabilityai/stable-diffusion-2" # generation time: 4 sec
+
+pipe = StableDiffusion3Pipeline.from_pretrained(model, torch_dtype=torch.float16)
+#pipe = StableDiffusionPipeline.from_pretrained(model, torch_dtype=torch.float16)
+pipe = pipe.to("cuda")
+
+model_loaded_timestamp = time.time()
+model_load_time = model_loaded_timestamp - start_timestamp
+print(f"model load time: {round(model_load_time)} seconds")
+
+prompt = "A majestic broadsword with a golden pommel, no background"
+image = pipe(
+    prompt,
+    guidance_scale=3.0,
+).images[0]
+
+image_name = "image7.png"
+image.save(f"images/{image_name}")
+
+generation_time = time.time() - model_loaded_timestamp
+print(f"image generation time: {round(generation_time)} seconds")