Update app.py

app.py

@@ -2,7 +2,6 @@ import gradio as gr
 from huggingface_hub import InferenceClient
 from transformers import (
     AutoProcessor, LlavaForConditionalGeneration,
-    IdeficsForConditionalGeneration,
     FuyuForCausalLM,
 )
 import torch
@@ -17,19 +16,13 @@ llava_model = LlavaForConditionalGeneration.from_pretrained(
     "llava-hf/llava-1.6-mistral-7b-hf", torch_dtype=torch.float16, device_map="auto"
 )
 
-# === 3. IDEFICS ===
-idefics_proc = AutoProcessor.from_pretrained("HuggingFaceM4/idefics-9b-instruct")
-idefics_model = IdeficsForConditionalGeneration.from_pretrained(
-    "HuggingFaceM4/idefics-9b-instruct", torch_dtype=torch.float16, device_map="auto"
-)
-
-# === 4. Fuyu ===
+# === 3. Fuyu ===
 fuyu_proc = AutoProcessor.from_pretrained("adept/fuyu-8b")
 fuyu_model = FuyuForCausalLM.from_pretrained(
     "adept/fuyu-8b", torch_dtype=torch.float16, device_map="auto"
 )
 
-# === 5. Gabungan Multi-Model Handler ===
+# === 4. Gabungan Multi-Model Handler ===
 def smart_respond(message, history, image=None):
     if image:
         results = []
@@ -43,16 +36,6 @@ def smart_respond(message, history, image=None):
         except Exception as e:
             results.append(f"LLaVA error: {e}")
 
-        # — IDEFICS
-        try:
-            prompt = [{"role": "user", "content": [{"type": "image"}, {"type": "text", "text": message}]}]
-            inputs = idefics_proc(prompt, images=[image], return_tensors="pt").to("cuda")
-            output = idefics_model.generate(**inputs, max_new_tokens=512)
-            idefics_reply = idefics_proc.decode(output[0], skip_special_tokens=True)
-            results.append(f"📄 **IDEFICS**:\n{idefics_reply}")
-        except Exception as e:
-            results.append(f"IDEFICS error: {e}")
-
         # — Fuyu
         try:
             fuyu_inputs = fuyu_proc(images=image, text=message, return_tensors="pt").to("cuda")
@@ -78,9 +61,9 @@ def smart_respond(message, history, image=None):
             response += token
             yield response
 
-# === 6. Gradio Interface (ChatGPT-style) ===
+# === 5. Gradio Interface (ChatGPT-style) ===
 with gr.Blocks() as demo:
-    gr.Markdown("## 🤖 Elaina AI — Teks + Gambar (4 Model)")
+    gr.Markdown("## 🤖 Elaina AI — Teks + Gambar (LLaVA + Fuyu + OpenChat)")
     chatbot = gr.Chatbot()
     state = gr.State([])
 
@@ -104,6 +87,6 @@ with gr.Blocks() as demo:
        .then(bot_response, state, chatbot) \
        .then(lambda x: x, chatbot, state)
 
-# === 7. Jalankan ===
+# === 6. Jalankan ===
 if __name__ == "__main__":
     demo.launch()