Update app.py
Browse files
app.py
CHANGED
|
@@ -3,53 +3,21 @@ from llama_cpp import Llama
|
|
| 3 |
from huggingface_hub import hf_hub_download
|
| 4 |
|
| 5 |
print("Loading model... please wait.")
|
|
|
|
| 6 |
model_path = hf_hub_download(
|
| 7 |
repo_id="mradermacher/Qwen2.5-Coder-32B-Instruct-GGUF",
|
| 8 |
-
filename="Qwen2.5-Coder-32B-Instruct.
|
| 9 |
-
)
|
| 10 |
-
|
| 11 |
-
llm = Llama(
|
| 12 |
-
model_path=model_path,
|
| 13 |
-
n_ctx=2048,
|
| 14 |
-
n_threads=2
|
| 15 |
)
|
| 16 |
|
| 17 |
-
|
| 18 |
-
formatted_messages = [{"role": "system", "content": system_message}]
|
| 19 |
-
for user_msg, assistant_msg in history:
|
| 20 |
-
formatted_messages.append({"role": "user", "content": user_msg})
|
| 21 |
-
formatted_messages.append({"role": "assistant", "content": assistant_msg})
|
| 22 |
-
formatted_messages.append({"role": "user", "content": message})
|
| 23 |
|
| 24 |
-
|
|
|
|
| 25 |
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
max_tokens=max_tokens,
|
| 29 |
-
temperature=temperature,
|
| 30 |
-
top_p=top_p,
|
| 31 |
-
stream=True
|
| 32 |
-
)
|
| 33 |
|
| 34 |
-
|
| 35 |
-
if "content" in chunk["choices"][0]["delta"]:
|
| 36 |
-
token = chunk["choices"][0]["delta"]["content"]
|
| 37 |
-
response_text += token
|
| 38 |
-
yield response_text
|
| 39 |
-
|
| 40 |
-
# --- Gradio UI Setup ---
|
| 41 |
-
with gr.Blocks() as demo:
|
| 42 |
-
gr.Markdown("Qwen2.5-Coder-32B (2-bit)")
|
| 43 |
-
|
| 44 |
-
chatbot = gr.ChatInterface(
|
| 45 |
-
respond,
|
| 46 |
-
additional_inputs=[
|
| 47 |
-
gr.Textbox(value="You are an Elite Minecraft Modder.", label="System message"),
|
| 48 |
-
gr.Slider(minimum=1, maximum=1024, value=512, step=1, label="Max new tokens"),
|
| 49 |
-
gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.1, label="Temperature"),
|
| 50 |
-
gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.05, label="Top-p"),
|
| 51 |
-
],
|
| 52 |
-
)
|
| 53 |
|
| 54 |
if __name__ == "__main__":
|
| 55 |
demo.launch()
|
|
|
|
| 3 |
from huggingface_hub import hf_hub_download
|
| 4 |
|
| 5 |
print("Loading model... please wait.")
|
| 6 |
+
|
| 7 |
model_path = hf_hub_download(
|
| 8 |
repo_id="mradermacher/Qwen2.5-Coder-32B-Instruct-GGUF",
|
| 9 |
+
filename="Qwen2.5-Coder-32B-Instruct.Q2_K.gguf"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
)
|
| 11 |
|
| 12 |
+
llm = Llama(model_path=model_path, n_ctx=2048)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
|
| 14 |
+
def respond(message, history):
|
| 15 |
+
prompt = f"System: You are a helpful AI.\nUser: {message}\nAssistant:"
|
| 16 |
|
| 17 |
+
output = llm(prompt, max_tokens=512, stop=["User:"], echo=False)
|
| 18 |
+
return output["choices"][0]["text"]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
+
demo = gr.ChatInterface(fn=respond, title="Qwen 32B Coder (CPU)")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
if __name__ == "__main__":
|
| 23 |
demo.launch()
|