Update inference.py

inference.py

@@ -1,61 +1,30 @@
-import os
-
-# ===== CPU optimization =====
-CPU_THREADS = str(os.cpu_count() or 2)
-
-os.environ["OMP_NUM_THREADS"] = CPU_THREADS
-os.environ["OPENVINO_NUM_THREADS"] = CPU_THREADS
-os.environ["MKL_NUM_THREADS"] = CPU_THREADS
-os.environ["NUMEXPR_NUM_THREADS"] = CPU_THREADS
-os.environ["OMP_WAIT_POLICY"] = "PASSIVE"
-os.environ["KMP_BLOCKTIME"] = "0"
-
 import torch
 from optimum.intel import OVZImagePipeline
 
-# ===== Global =====
-pipe = None
-generator = torch.Generator("cpu")
-
-
-def load_model():
-    global pipe
-
-    if pipe is None:
-        pipe = OVZImagePipeline.from_pretrained(
-            "hsuwill000/Z-Image-Turbo-ov",
-            device="cpu"
-        )
-
-        # warmup
-        pipe(
-            prompt="warmup",
-            height=512,
-            width=512,
-            num_inference_steps=1,
-            guidance_scale=0.0,
-        )
-
-    return pipe
+# ===== Load once =====
+pipe = OVZImagePipeline.from_pretrained(
+    "hsuwill000/Z-Image-Turbo-ov",
+    device="cpu"
+)
 
 
 def generate(
-    prompt: str,
-    height: int = 512,
-    width: int = 512,
-    steps: int = 9,
-    seed: int = -1,
+    prompt,
+    height=512,
+    width=512,
+    steps=9,
+    seed=-1,
 ):
-    pipe = load_model()
-
-    if seed != -1:
-        generator.manual_seed(int(seed))
+    if seed == -1:
+        generator = torch.Generator("cpu")
+    else:
+        generator = torch.Generator("cpu").manual_seed(int(seed))
 
     image = pipe(
         prompt=prompt,
-        height=height,
-        width=width,
-        num_inference_steps=steps,
+        height=int(height),
+        width=int(width),
+        num_inference_steps=int(steps),
         guidance_scale=0.0,
         generator=generator,
     ).images[0]