Upload 2 files
Browse files- app.py +23 -0
- requirements.txt +2 -0
app.py
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from flask import Flask, request, jsonify
|
| 2 |
+
from llama_cpp import Llama
|
| 3 |
+
|
| 4 |
+
app = Flask(__name__)
|
| 5 |
+
|
| 6 |
+
# Load Maira - using the exact filename you downloaded
|
| 7 |
+
llm = Llama(model_path="SmolLM2-360M-Instruct.Q4_K_M.gguf", n_ctx=2048)
|
| 8 |
+
|
| 9 |
+
@app.route('/chat', methods=['POST'])
|
| 10 |
+
def chat():
|
| 11 |
+
data = request.json
|
| 12 |
+
user_input = data.get("message", "")
|
| 13 |
+
|
| 14 |
+
# The format she learned in training
|
| 15 |
+
prompt = f"### User: {user_input}\n### Maira:"
|
| 16 |
+
|
| 17 |
+
output = llm(prompt, max_tokens=150, stop=["###", "</s>"], echo=False)
|
| 18 |
+
response = output["choices"][0]["text"].strip()
|
| 19 |
+
|
| 20 |
+
return jsonify({"maira": response})
|
| 21 |
+
|
| 22 |
+
if __name__ == "__main__":
|
| 23 |
+
app.run(host="0.0.0.0", port=10000)
|
requirements.txt
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
flask
|
| 2 |
+
llama-cpp-python
|