Upload 4 files

app.py

@@ -1,70 +1,70 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+import subprocess
+from pathlib import Path
 
+# Caminhos dos modelos
+MODEL_DIR = Path("./models")
+MODELS = {
+    "Q3_K_XL (Inferior)": MODEL_DIR / "Llama-3.2-1B-Instruct-Q3_K_XL.gguf",
+    "Q4_K_M (Alta)": MODEL_DIR / "Llama-3.2-1B-Instruct-Q4_K_M.gguf",
+}
 
-def respond(
-    message,
-    history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
-):
+# Histórico global de chat
+history_dict = {}
+
+def run_llama(prompt, model_path, max_tokens, temperature):
     """
-    For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
+    Executa o llama.cpp localmente e retorna a saída
     """
-    client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")
-
-    messages = [{"role": "system", "content": system_message}]
-
-    messages.extend(history)
+    cmd = [
+        "./llama.cpp/main",
+        "-m", str(model_path),
+        "-p", prompt,
+        "-n", str(max_tokens),
+        "-t", "4",
+        "-temp", str(temperature),
+        "--color", "false",
+    ]
 
-    messages.append({"role": "user", "content": message})
+    result = subprocess.run(cmd, capture_output=True, text=True)
+    return result.stdout.strip()
 
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
+def respond(user_message, model_choice, max_tokens, temperature, session_id="default"):
+    """
+    Função chamada pelo Gradio, mantém histórico
+    """
+    global history_dict
+    if session_id not in history_dict:
+        history_dict[session_id] = []
 
-        response += token
-        yield response
+    history = history_dict[session_id]
+    # Concatena histórico em um único prompt
+    prompt = "\n".join([f"User: {m['user']}\nAI: {m['ai']}" for m in history])
+    prompt += f"\nUser: {user_message}\nAI:"
 
+    # Roda o llama.cpp
+    response = run_llama(prompt, MODELS[model_choice], max_tokens, temperature)
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-chatbot = gr.ChatInterface(
-    respond,
-    type="messages",
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
+    # Atualiza histórico
+    history.append({"user": user_message, "ai": response})
+    return response
 
+# Interface Gradio
 with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
-    chatbot.render()
+    with gr.Row():
+        model_dropdown = gr.Dropdown(list(MODELS.keys()), label="Escolha o modelo")
+        user_input = gr.Textbox(label="Mensagem do usuário")
+    with gr.Row():
+        max_tokens_slider = gr.Slider(1, 2048, value=512, step=1, label="Max tokens")
+        temperature_slider = gr.Slider(0.1, 4.0, value=0.7, step=0.1, label="Temperature")
+    output_box = gr.Textbox(label="Resposta")
 
+    btn = gr.Button("Enviar")
+    btn.click(
+        respond,
+        inputs=[user_input, model_dropdown, max_tokens_slider, temperature_slider],
+        outputs=[output_box],
+    )
 
 if __name__ == "__main__":
     demo.launch()

download_models.py

@@ -0,0 +1,24 @@
+from huggingface_hub import hf_hub_download
+from pathlib import Path
+
+# Cria pasta models se não existir
+MODEL_DIR = Path("./models")
+MODEL_DIR.mkdir(exist_ok=True)
+
+# Dicionário com os modelos e os arquivos GGUF correspondentes
+MODELS = {
+    "Q3_K_XL": "Llama-3.2-1B-Instruct-Q3_K_XL.gguf",
+    "Q4_K_M": "Llama-3.2-1B-Instruct-Q4_K_M.gguf",
+}
+
+REPO_ID = "bartowski/Llama-3.2-1B-Instruct-GGUF"
+
+for key, file_name in MODELS.items():
+    print(f"Baixando {file_name}...")
+    path = hf_hub_download(
+        repo_id=REPO_ID,
+        filename=file_name,
+        local_dir=MODEL_DIR,
+        force_filename=file_name
+    )
+    print(f"{file_name} salvo em {path}")

install.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+# Compila o llama.cpp no Space
+cd llama.cpp
+make
+cd ..

requirements.txt

@@ -0,0 +1 @@
+gradio