fix stuff

app.py

@@ -5,17 +5,24 @@ MAX_TOKENS = 64
 
 llm = Llama(model_path="ggml-model-f16-q4_0.bin", n_ctx=2048)
 
-def generate_text(prompt):
-    output = llm(prompt, max_tokens=MAX_TOKENS, echo=False, stop=["</s>", "\n", "User:"])
-    output_text = output['choices'][0]['text']
-    return output_text
+def generate_text_instruct(input_text):
+    response = ""
+    txt2tag_prompt = f"Provide me with danbooru tags that accurately describe the following description. {input_text}"
+    for output in llm(f"A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. USER: {txt2tag_prompt} ASSISTANT: ", 
+                      echo=False, stream=True, stop=["</s>", "\n", "User:"]):
+        answer = output['choices'][0]['text']
+        response += answer
+        yield response
 
-description = f"txt2tag 13B base model, max_tokens={MAX_TOKENS}"
-
-gradio_interface = gr.Interface(
-  fn=generate_text,
-  inputs="text",
-  outputs="text",
-  title="txt2tag API",
+instruct_interface = gr.Interface(
+    fn=generate_text_instruct,
+    inputs=gr.inputs.Textbox(lines= 10, label="Enter your instruction text"),
+    outputs=gr.outputs.Textbox(label="Output text"),
 )
-gradio_interface.launch()
+
+with gr.Blocks() as demo:
+    with gr.Tab("Instruct"):
+        gr.Markdown("# GGML Spaces Instruct Demo")
+        instruct_interface.render()
+
+demo.queue(max_size=16, concurrency_count=1).launch(debug=True)