Update app.py

app.py

@@ -1,29 +1,48 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
-import os
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
 
-token = os.environ.get("HF_TOKEN")
+model_name = "Qwen/Qwen2.5-Coder-1.5B-Instruct"
 
-client = InferenceClient(
-    model="HuggingFaceH4/zephyr-7b-beta",
-    token=token
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    dtype=torch.float32,
+    low_cpu_mem_usage=True
 )
+model.eval()
 
 def generate(prompt):
-    try:
-        response = client.chat_completion(
-            messages=[{"role": "user", "content": prompt}],
-            max_tokens=512,
-            temperature=0.7
+    formatted = f"### Instruction:\n{prompt}\n### Response:\n"
+    
+    inputs = tokenizer(
+        formatted,
+        return_tensors="pt",
+        truncation=True,
+        max_length=512
+    )
+    
+    input_ids = inputs["input_ids"]
+    attention_mask = inputs["attention_mask"]
+    
+    with torch.no_grad():
+        output = model.generate(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            max_new_tokens=256,
+            do_sample=True,
+            temperature=0.7,
+            pad_token_id=tokenizer.eos_token_id
         )
-        return response.choices[0].message.content
-    except Exception as e:
-        return f"Error: {str(e)}"
+    
+    new_tokens = output[0][input_ids.shape[-1]:]
+    result = tokenizer.decode(new_tokens, skip_special_tokens=True)
+    return result.strip()
 
 gr.Interface(
     fn=generate,
     inputs=gr.Textbox(lines=4, label="Tanya apa saja"),
     outputs=gr.Textbox(label="Jawaban"),
     title="AI Assistant",
-    description="Powered by HF Inference API"
+    description="Powered by Qwen2.5-Coder 1.5B"
 ).launch()