新增 .gitignore 檔案並重構 app.py，整合 Stable Diffusion 與 BLIP 功能

.gitignore

@@ -0,0 +1 @@
+venv/

app.py

@@ -1,154 +1,183 @@
-import gradio as gr
-import numpy as np
-import random
-
-# import spaces #[uncomment to use ZeroGPU]
-from diffusers import DiffusionPipeline
+import os
 import torch
+import gradio as gr
+from PIL import Image
+from diffusers import StableDiffusionPipeline
+from transformers import BlipProcessor, BlipForConditionalGeneration
+
+# -----------------------------
+# 環境與效能設定（CPU 友善）
+# -----------------------------
+os.environ.setdefault("HF_HUB_DISABLE_TELEMETRY", "1")
+os.environ.setdefault("TOKENIZERS_PARALLELISM", "false")
+
+# 限制 CPU 執行緒數（避免在 Spaces 上搶太多資源）
+torch.set_num_threads(max(1, min(4, os.cpu_count() or 2)))
+
+DEVICE = "cpu"  # 若有 GPU 可改為 "cuda"
+DTYPE = torch.float32
+
+# -----------------------------
+# 載入模型（首次啟動會自動下載）
+# -----------------------------
+
+def load_sd_pipe():
+    pipe = StableDiffusionPipeline.from_pretrained(
+        "runwayml/stable-diffusion-v1-5",
+        torch_dtype=DTYPE,
+        safety_checker=None,  # 若課堂需要內容審查可改回預設 safety_checker
+    )
+    pipe = pipe.to(DEVICE)
+    # 省記憶體設定（對 CPU/低資源環境友好）
+    pipe.enable_attention_slicing()
+    pipe.enable_vae_tiling()
+    return pipe
+
+
+def load_blip():
+    processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
+    model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base").to(DEVICE)
+    return processor, model
+
+
+SD_PIPE = load_sd_pipe()
+BLIP_PROCESSOR, BLIP_MODEL = load_blip()
+
+
+# -----------------------------
+# 功能：Text → Image（Stable Diffusion）
+# -----------------------------
+
+def txt2img(prompt: str, neg_prompt: str, steps: int, guidance: float, seed: int | None):
+    if not prompt or prompt.strip() == "":
+        return None, "請輸入 prompt。"
+
+    generator = None
+    if seed is not None and seed >= 0:
+        generator = torch.Generator(device=DEVICE).manual_seed(int(seed))
+
+    with torch.inference_mode():
+        image = SD_PIPE(
+            prompt=prompt,
+            negative_prompt=neg_prompt or None,
+            num_inference_steps=int(steps),
+            guidance_scale=float(guidance),
+            generator=generator,
+            height=512,
+            width=512,
+        ).images[0]
+    return image, "生成完成"
+
+
+# -----------------------------
+# 功能：Image → Text（BLIP Caption）
+# -----------------------------
+
+def caption_image(image: Image.Image, max_len: int = 50):
+    if image is None:
+        return "請先提供圖片"
+
+    image = image.convert("RGB")
+    inputs = BLIP_PROCESSOR(image, return_tensors="pt").to(DEVICE)
+    with torch.inference_mode():
+        out = BLIP_MODEL.generate(**inputs, max_length=int(max_len))
+    caption = BLIP_PROCESSOR.decode(out[0], skip_special_tokens=True)
+    return caption
+
+
+# -----------------------------
+# 功能：一鍵串接（Text → Image → Caption）
+# -----------------------------
+
+def generate_and_caption(prompt: str, neg_prompt: str, steps: int, guidance: float, seed: int | None, max_len: int):
+    image, _ = txt2img(prompt, neg_prompt, steps, guidance, seed)
+    if image is None:
+        return None, ""
+    cap = caption_image(image, max_len=max_len)
+    return image, cap
+
+
+# -----------------------------
+# Gradio 介面
+# -----------------------------
+with gr.Blocks(title="Stable Diffusion + BLIP", theme=gr.themes.Soft()) as demo:
+    gr.Markdown(
+        """
+        # 🖼️ Stable Diffusion + 📝 BLIP
+        一個簡單的 Vision-Language 展示：**文字生圖** 與 **圖片描述**。\
+        在 CPU 上執行可能較慢，建議把步數調低（例如 15）。
+        """
+    )
 
-device = "cuda" if torch.cuda.is_available() else "cpu"
-model_repo_id = "stabilityai/sdxl-turbo"  # Replace to the model you would like to use
-
-if torch.cuda.is_available():
-    torch_dtype = torch.float16
-else:
-    torch_dtype = torch.float32
-
-pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
-pipe = pipe.to(device)
-
-MAX_SEED = np.iinfo(np.int32).max
-MAX_IMAGE_SIZE = 1024
-
-
-# @spaces.GPU #[uncomment to use ZeroGPU]
-def infer(
-    prompt,
-    negative_prompt,
-    seed,
-    randomize_seed,
-    width,
-    height,
-    guidance_scale,
-    num_inference_steps,
-    progress=gr.Progress(track_tqdm=True),
-):
-    if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-
-    generator = torch.Generator().manual_seed(seed)
-
-    image = pipe(
-        prompt=prompt,
-        negative_prompt=negative_prompt,
-        guidance_scale=guidance_scale,
-        num_inference_steps=num_inference_steps,
-        width=width,
-        height=height,
-        generator=generator,
-    ).images[0]
-
-    return image, seed
-
-
-examples = [
-    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
-    "An astronaut riding a green horse",
-    "A delicious ceviche cheesecake slice",
-]
-
-css = """
-#col-container {
-    margin: 0 auto;
-    max-width: 640px;
-}
-"""
-
-with gr.Blocks(css=css) as demo:
-    with gr.Column(elem_id="col-container"):
-        gr.Markdown(" # Text-to-Image Gradio Template")
-
-        with gr.Row():
-            prompt = gr.Text(
-                label="Prompt",
-                show_label=False,
-                max_lines=1,
-                placeholder="Enter your prompt",
-                container=False,
-            )
-
-            run_button = gr.Button("Run", scale=0, variant="primary")
-
-        result = gr.Image(label="Result", show_label=False)
-
-        with gr.Accordion("Advanced Settings", open=False):
-            negative_prompt = gr.Text(
-                label="Negative prompt",
-                max_lines=1,
-                placeholder="Enter a negative prompt",
-                visible=False,
-            )
-
-            seed = gr.Slider(
-                label="Seed",
-                minimum=0,
-                maximum=MAX_SEED,
-                step=1,
-                value=0,
-            )
-
-            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-
+    with gr.Tabs():
+        # Tab 1: Text -> Image
+        with gr.TabItem("Text → Image"):
             with gr.Row():
-                width = gr.Slider(
-                    label="Width",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024,  # Replace with defaults that work for your model
-                )
-
-                height = gr.Slider(
-                    label="Height",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024,  # Replace with defaults that work for your model
-                )
+                with gr.Column(scale=1):
+                    prompt = gr.Textbox(label="Prompt", placeholder="a cozy cat reading a book by the window, warm lighting")
+                    neg_prompt = gr.Textbox(label="Negative Prompt", placeholder="blurry, low quality, watermark", value="")
+                    steps = gr.Slider(5, 50, value=15, step=1, label="Steps (越高越慢)")
+                    guidance = gr.Slider(1.0, 12.0, value=7.5, step=0.5, label="Guidance Scale")
+                    seed = gr.Number(label="Seed（相同設定可重現，-1 表示隨機）", value=-1, precision=0)
+                    run_btn = gr.Button("🚀 生成圖片", variant="primary")
+                with gr.Column(scale=1):
+                    out_img = gr.Image(label="Generated Image", format="png")
+                    status = gr.Markdown()
+
+            def _run_txt2img(prompt, neg_prompt, steps, guidance, seed):
+                with gr.Progress(track_tqdm=True) as prog:
+                    prog(0, desc="準備模型…")
+                    image, msg = txt2img(prompt, neg_prompt, steps, guidance, int(seed) if seed is not None else None)
+                    prog(1, desc="完成")
+                    return image, msg
+
+            run_btn.click(_run_txt2img, [prompt, neg_prompt, steps, guidance, seed], [out_img, status])
+
+        # Tab 2: Image -> Caption
+        with gr.TabItem("Image → Caption"):
+            with gr.Row():
+                with gr.Column(scale=1):
+                    in_img = gr.Image(label="Upload Image", type="pil")
+                    max_len = gr.Slider(10, 100, value=50, step=5, label="Caption 長度上限")
+                    cap_btn = gr.Button("📝 產生描述")
+                with gr.Column(scale=1):
+                    caption_out = gr.Textbox(label="BLIP Caption")
+
+            cap_btn.click(caption_image, [in_img, max_len], [caption_out])
 
+        # Tab 3: 一鍵串接
+        with gr.TabItem("Generate → Caption"):
             with gr.Row():
-                guidance_scale = gr.Slider(
-                    label="Guidance scale",
-                    minimum=0.0,
-                    maximum=10.0,
-                    step=0.1,
-                    value=0.0,  # Replace with defaults that work for your model
-                )
-
-                num_inference_steps = gr.Slider(
-                    label="Number of inference steps",
-                    minimum=1,
-                    maximum=50,
-                    step=1,
-                    value=2,  # Replace with defaults that work for your model
-                )
-
-        gr.Examples(examples=examples, inputs=[prompt])
-    gr.on(
-        triggers=[run_button.click, prompt.submit],
-        fn=infer,
-        inputs=[
-            prompt,
-            negative_prompt,
-            seed,
-            randomize_seed,
-            width,
-            height,
-            guidance_scale,
-            num_inference_steps,
+                with gr.Column(scale=1):
+                    g_prompt = gr.Textbox(label="Prompt", placeholder="a cozy cat reading a book by the window, warm lighting")
+                    g_neg_prompt = gr.Textbox(label="Negative Prompt", value="")
+                    g_steps = gr.Slider(5, 50, value=15, step=1, label="Steps")
+                    g_guidance = gr.Slider(1.0, 12.0, value=7.5, step=0.5, label="Guidance Scale")
+                    g_seed = gr.Number(label="Seed（-1 為隨機）", value=-1, precision=0)
+                    g_max_len = gr.Slider(10, 100, value=50, step=5, label="Caption 長度上限")
+                    g_btn = gr.Button("✨ 一鍵生成 + 描述", variant="primary")
+                with gr.Column(scale=1):
+                    g_img = gr.Image(label="Generated Image", format="png")
+                    g_cap = gr.Textbox(label="BLIP Caption")
+
+            def _run_generate_and_caption(prompt, neg_prompt, steps, guidance, seed, max_len):
+                with gr.Progress(track_tqdm=True) as prog:
+                    prog(0, desc="生成圖片中…")
+                    image, caption = generate_and_caption(prompt, neg_prompt, steps, guidance, int(seed) if seed is not None else None, int(max_len))
+                    prog(1, desc="完成")
+                    return image, caption
+
+            g_btn.click(_run_generate_and_caption, [g_prompt, g_neg_prompt, g_steps, g_guidance, g_seed, g_max_len], [g_img, g_cap])
+
+    gr.Examples(
+        examples=[
+            ["a cozy cat reading a book by the window, warm lighting", "", 15, 7.5, 42],
+            ["cinematic portrait of an astronaut in a forest, film grain, bokeh", "", 20, 8.0, 1234],
         ],
-        outputs=[result, seed],
+        inputs=[prompt, neg_prompt, steps, guidance, seed],
+        label="Prompt 範例（可直接點選）",
     )
 
 if __name__ == "__main__":
-    demo.launch()
+    # 在 HF Spaces 使用：不用 demo.launch(share=True)
+    demo.launch()