Update app.py

app.py

@@ -1,18 +1,24 @@
 import gradio as gr
-from transformers import pipeline, WhisperProcessor, WhisperForConditionalGeneration
-from diffusers import AutoPipelineForText2Image
+from transformers import pipeline, T5ForConditionalGeneration, T5Tokenizer
 import torch
+import stable_diffusion_webnn  # 假设stable-diffusion-v1.5-webnn的库名为 stable_diffusion_webnn
 
-# 使用BART模型生成文本描述
-prompt_generator = pipeline("text2text-generation", model="facebook/bart-large-cnn")
+# 使用T5模型生成文本描述
+model_name = "t5-large"  # 可以根据需求选择不同版本的T5
+tokenizer = T5Tokenizer.from_pretrained(model_name)
+t5_model = T5ForConditionalGeneration.from_pretrained(model_name)
 
 def generate_prompt(description: str) -> str:
-    # 根据简短描述生成详细的图像生成提示
-    prompt = prompt_generator(f"将这个描述扩展为一个详细的图像生成提示：{description}", max_length=150)[0]['generated_text']
+    # 使用T5模型生成详细的图像生成提示
+    input_text = f"将这个描述扩展为一个详细的图像生成提示：{description}"
+    inputs = tokenizer(input_text, return_tensors="pt", max_length=512, truncation=True)
+    outputs = t5_model.generate(inputs["input_ids"], max_length=150, num_beams=5, early_stopping=True)
+    prompt = tokenizer.decode(outputs[0], skip_special_tokens=True)
     return prompt
 
-# 使用AutoPipelineForText2Image替换StableDiffusionPipeline
-text2image_pipeline = AutoPipelineForText2Image.from_pretrained("stabilityai/stable-diffusion-2-1-base")
+# 使用 stable-diffusion-v1.5-webnn 库加载 Stable Diffusion 模型
+# 这里假设 stable_diffusion_webnn 可以直接加载模型并生成图像
+text2image_pipeline = stable_diffusion_webnn.StableDiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-2-1-base")
 text2image_pipeline.to("cpu")  # 使用CPU
 
 def generate_image(prompt: str):
@@ -21,6 +27,7 @@ def generate_image(prompt: str):
     return image
 
 # 使用Whisper模型进行语音转文本
+from transformers import WhisperProcessor, WhisperForConditionalGeneration
 processor = WhisperProcessor.from_pretrained("openai/whisper-large")
 model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-large")