Spaces:
Runtime error
Runtime error
Commit
·
6a4f05e
0
Parent(s):
premier commit
Browse files- .gitignore +6 -0
- app.py +32 -0
- main.py +10 -0
.gitignore
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Fichiers et dossiers à ignorer
|
| 2 |
+
llamavenv/
|
| 3 |
+
models/
|
| 4 |
+
*.gguf
|
| 5 |
+
__pycache__/
|
| 6 |
+
*.pyc
|
app.py
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
from llama_cpp import Llama
|
| 3 |
+
import os
|
| 4 |
+
|
| 5 |
+
st.set_page_config(page_title="Chatbot RAG local",page_icon="🤖")
|
| 6 |
+
|
| 7 |
+
@st.cache_resource
|
| 8 |
+
def load_model():
|
| 9 |
+
model_path="models/phi-2.Q4_K_M.gguf"
|
| 10 |
+
if not os.path.exists(model_path):
|
| 11 |
+
raise FileNotFoundError(f"Modèle non trouvé: {model_path}")
|
| 12 |
+
return Llama(model_path=model_path,n_ctx=2048,n_threads=4)
|
| 13 |
+
|
| 14 |
+
llm=load_model()
|
| 15 |
+
|
| 16 |
+
st.title("🤖 Chatbot LLM Local (CPU)")
|
| 17 |
+
|
| 18 |
+
user_input=st.text_area("Posez votre question :", height=100)
|
| 19 |
+
|
| 20 |
+
if st.button("Envoyer") and user_input.strip():
|
| 21 |
+
with st.spinner("Génération en cours..."):
|
| 22 |
+
full_prompt = f"### Instruction: {user_input.strip()}\n### Réponse:"
|
| 23 |
+
#full_prompt = f"Question: {user_input.strip()}\nAnswer:"
|
| 24 |
+
#output = llm(full_prompt, max_tokens=100, stop=["Question:", "Answer:", "\n\n"])
|
| 25 |
+
output = llm(full_prompt, max_tokens=150, stop=["### Instruction:"])
|
| 26 |
+
#output = llm(full_prompt, max_tokens=80)
|
| 27 |
+
#response = output["choices"][0]["text"]
|
| 28 |
+
response = output["choices"][0]["text"].strip()
|
| 29 |
+
response = response.split("### Instruction:")[0].strip()
|
| 30 |
+
|
| 31 |
+
st.markdown("**Réponse :**")
|
| 32 |
+
st.success(response)
|
main.py
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from llama_cpp import Llama
|
| 2 |
+
|
| 3 |
+
llm = Llama(
|
| 4 |
+
model_path="models/Nous-Hermes-2-Mistral-7B-DPO.Q4_K_M.gguf", # ajuste le chemin si nécessaire
|
| 5 |
+
n_ctx=2048,
|
| 6 |
+
n_threads=4
|
| 7 |
+
)
|
| 8 |
+
|
| 9 |
+
response = llm("### Instruction: Quelle est la capitale du Sénégal ?\n### Réponse:", max_tokens=128)
|
| 10 |
+
print(response["choices"][0]["text"])
|