TF

MMaDA/inference/gradio_multimodal_demo_inst.py

@@ -467,33 +467,6 @@ def _load_t2i_examples():
     return [[line] for line in lines]
 
 
-def _load_i2i_examples():
-    d = ASSET_ROOT / "i2i"
-    if not d.exists():
-        return []
-
-    # 이미지 파일들 (image1.jpeg, image2.png, ...)
-    image_files = sorted(
-        [p for p in d.iterdir() if p.suffix.lower() in {".png", ".jpg", ".jpeg", ".webp"}]
-    )
-    # 텍스트 파일들 (text1.txt, text2.txt, ...)
-    text_files = sorted(
-        [p for p in d.iterdir() if p.suffix.lower() == ".txt"]
-    )
-
-    n = min(len(image_files), len(text_files))
-    examples = []
-    for i in range(n):
-        img_path = image_files[i]
-        txt_path = text_files[i]
-        instruction = txt_path.read_text(encoding="utf-8").strip()
-        if not instruction:
-            continue
-        # Gradio Examples 형식: [image, instruction_text]
-        examples.append([str(img_path), instruction])
-    return examples
-
-
 def _load_media_examples(subdir: str, suffixes):
     target_dir = DEMO_ROOT / subdir
     if not target_dir.exists():

app.py

@@ -182,11 +182,37 @@ def _load_media_examples(subdir: str, suffixes):
             ex.append([str(p)])
     return ex
 
+def _load_i2i_examples():
+    d = ASSET_ROOT / "i2i"
+    if not d.exists():
+        return []
+
+    # 이미지 파일들 (image1.jpeg, image2.png, ...)
+    image_files = sorted(
+        [p for p in d.iterdir() if p.suffix.lower() in {".png", ".jpg", ".jpeg", ".webp"}]
+    )
+    # 텍스트 파일들 (text1.txt, text2.txt, ...)
+    text_files = sorted(
+        [p for p in d.iterdir() if p.suffix.lower() == ".txt"]
+    )
+
+    n = min(len(image_files), len(text_files))
+    examples = []
+    for i in range(n):
+        img_path = image_files[i]
+        txt_path = text_files[i]
+        instruction = txt_path.read_text(encoding="utf-8").strip()
+        if not instruction:
+            continue
+        # Gradio Examples 형식: [image, instruction_text]
+        examples.append([str(img_path), instruction])
+    return examples
 
 # text-based examples
 T2S_EXAMPLES = _load_text_examples(ASSET_ROOT / "t2s" / "text.txt")
 CHAT_EXAMPLES = _load_text_examples(ASSET_ROOT / "chat" / "text.txt")
 T2I_EXAMPLES = _load_text_examples(ASSET_ROOT / "t2i" / "text.txt")
+I2I_EXAMPLES = _load_i2i_examples()
 
 # audio / video / image examples
 S2T_EXAMPLES = _load_media_examples("s2t", {".wav", ".mp3", ".flac", ".ogg"})