Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -36,9 +36,9 @@ def load_model(model_key: str):
|
|
| 36 |
model_cache = {}
|
| 37 |
|
| 38 |
def respond(message: str,
|
| 39 |
-
history: List[
|
| 40 |
model_key: str,
|
| 41 |
-
system_prompt: str) -> Tuple[List[
|
| 42 |
"""Локальный чат с pipeline."""
|
| 43 |
try:
|
| 44 |
if model_key not in model_cache:
|
|
@@ -47,11 +47,16 @@ def respond(message: str,
|
|
| 47 |
|
| 48 |
print(f"🚀 Генерация: {model_key}, Msg='{message[:30]}...'")
|
| 49 |
|
| 50 |
-
#
|
| 51 |
messages = []
|
| 52 |
if system_prompt.strip():
|
| 53 |
messages.append({"role": "system", "content": system_prompt})
|
| 54 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
messages.append({"role": "user", "content": message})
|
| 56 |
|
| 57 |
# Apply chat template (для instruct)
|
|
@@ -64,38 +69,34 @@ def respond(message: str,
|
|
| 64 |
|
| 65 |
print(f"✅ Ответ: {bot_reply[:50]}...")
|
| 66 |
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
{"role": "assistant", "content": bot_reply}
|
| 70 |
-
]
|
| 71 |
return new_history, "", gr.update(value="")
|
| 72 |
|
| 73 |
except Exception as e:
|
| 74 |
error_msg = f"❌ {model_key}: {str(e)}"
|
| 75 |
print(f"💥 {error_msg}")
|
| 76 |
-
new_history = history + [
|
| 77 |
-
{"role": "user", "content": message},
|
| 78 |
-
{"role": "assistant", "content": error_msg}
|
| 79 |
-
]
|
| 80 |
return new_history, error_msg, gr.update(value="")
|
| 81 |
|
| 82 |
-
# UI
|
| 83 |
with gr.Blocks(title="🚀 Локальный HF Чат (на слабом CPU!)") as demo:
|
| 84 |
gr.Markdown("# Локальный Inference (без API!)\n**Маленькие модели** — 1-3 сек CPU. Большие думают ооочень долго. Нет limits/token. В качестве примера.")
|
| 85 |
|
| 86 |
-
with gr.Row(
|
| 87 |
model_dropdown = gr.Dropdown(choices=list(MODELS.keys()), value="Qwen2.5-0.5B", label="🧠 Модель")
|
| 88 |
system_prompt = gr.Textbox(label="📝 System", placeholder="Ты весёлый ИИ.", lines=2)
|
| 89 |
|
| 90 |
-
|
|
|
|
| 91 |
|
| 92 |
with gr.Row():
|
| 93 |
-
msg_input = gr.Textbox(placeholder="Привет! (Enter)",
|
| 94 |
-
send_btn = gr.Button("📤", variant="primary"
|
| 95 |
|
| 96 |
with gr.Row():
|
| 97 |
-
clear_btn = gr.Button("🗑️
|
| 98 |
-
retry_btn = gr.Button("🔄
|
| 99 |
|
| 100 |
status = gr.Textbox(label="Логи", interactive=False, lines=4)
|
| 101 |
|
|
@@ -108,8 +109,8 @@ with gr.Blocks(title="🚀 Локальный HF Чат (на слабом CPU!)
|
|
| 108 |
clear_btn.click(clear, outputs=[chatbot, status, msg_input])
|
| 109 |
|
| 110 |
def retry(history):
|
| 111 |
-
if len(history) >=
|
| 112 |
-
return history[-
|
| 113 |
return ""
|
| 114 |
retry_btn.click(retry, inputs=[chatbot], outputs=[msg_input])
|
| 115 |
|
|
|
|
| 36 |
model_cache = {}
|
| 37 |
|
| 38 |
def respond(message: str,
|
| 39 |
+
history: List[List[str]],
|
| 40 |
model_key: str,
|
| 41 |
+
system_prompt: str) -> Tuple[List[List[str]], str, Dict[str, Any]]:
|
| 42 |
"""Локальный чат с pipeline."""
|
| 43 |
try:
|
| 44 |
if model_key not in model_cache:
|
|
|
|
| 47 |
|
| 48 |
print(f"🚀 Генерация: {model_key}, Msg='{message[:30]}...'")
|
| 49 |
|
| 50 |
+
# Преобразуем формат истории из Gradio в формат сообщений
|
| 51 |
messages = []
|
| 52 |
if system_prompt.strip():
|
| 53 |
messages.append({"role": "system", "content": system_prompt})
|
| 54 |
+
|
| 55 |
+
# Конвертируем историю из формата [["user_msg", "bot_msg"], ...]
|
| 56 |
+
for user_msg, bot_msg in history:
|
| 57 |
+
messages.append({"role": "user", "content": user_msg})
|
| 58 |
+
messages.append({"role": "assistant", "content": bot_msg})
|
| 59 |
+
|
| 60 |
messages.append({"role": "user", "content": message})
|
| 61 |
|
| 62 |
# Apply chat template (для instruct)
|
|
|
|
| 69 |
|
| 70 |
print(f"✅ Ответ: {bot_reply[:50]}...")
|
| 71 |
|
| 72 |
+
# Обновляем историю в формате Gradio
|
| 73 |
+
new_history = history + [[message, bot_reply]]
|
|
|
|
|
|
|
| 74 |
return new_history, "", gr.update(value="")
|
| 75 |
|
| 76 |
except Exception as e:
|
| 77 |
error_msg = f"❌ {model_key}: {str(e)}"
|
| 78 |
print(f"💥 {error_msg}")
|
| 79 |
+
new_history = history + [[message, error_msg]]
|
|
|
|
|
|
|
|
|
|
| 80 |
return new_history, error_msg, gr.update(value="")
|
| 81 |
|
| 82 |
+
# UI
|
| 83 |
with gr.Blocks(title="🚀 Локальный HF Чат (на слабом CPU!)") as demo:
|
| 84 |
gr.Markdown("# Локальный Inference (без API!)\n**Маленькие модели** — 1-3 сек CPU. Большие думают ооочень долго. Нет limits/token. В качестве примера.")
|
| 85 |
|
| 86 |
+
with gr.Row():
|
| 87 |
model_dropdown = gr.Dropdown(choices=list(MODELS.keys()), value="Qwen2.5-0.5B", label="🧠 Модель")
|
| 88 |
system_prompt = gr.Textbox(label="📝 System", placeholder="Ты весёлый ИИ.", lines=2)
|
| 89 |
|
| 90 |
+
# Убрал параметр type="messages" для совместимости со старой версией
|
| 91 |
+
chatbot = gr.Chatbot(height=500)
|
| 92 |
|
| 93 |
with gr.Row():
|
| 94 |
+
msg_input = gr.Textbox(placeholder="Привет! (Enter)", show_label=False, lines=1)
|
| 95 |
+
send_btn = gr.Button("📤 Отправить", variant="primary")
|
| 96 |
|
| 97 |
with gr.Row():
|
| 98 |
+
clear_btn = gr.Button("🗑️ Очистить")
|
| 99 |
+
retry_btn = gr.Button("🔄 Повторить")
|
| 100 |
|
| 101 |
status = gr.Textbox(label="Логи", interactive=False, lines=4)
|
| 102 |
|
|
|
|
| 109 |
clear_btn.click(clear, outputs=[chatbot, status, msg_input])
|
| 110 |
|
| 111 |
def retry(history):
|
| 112 |
+
if len(history) >= 1:
|
| 113 |
+
return history[-1][0] # Последнее сообщение пользователя
|
| 114 |
return ""
|
| 115 |
retry_btn.click(retry, inputs=[chatbot], outputs=[msg_input])
|
| 116 |
|