Spaces:
Runtime error
Runtime error
Added knowledge bases and simple query with top 5 results as a table.
Browse files- app.py +49 -2
- data/knowledge_base.parquet +3 -0
- data/knowledge_base_embeddings.pkl +3 -0
app.py
CHANGED
|
@@ -1,4 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import streamlit as st
|
|
|
|
| 2 |
|
| 3 |
-
|
| 4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import faiss
|
| 2 |
+
import joblib
|
| 3 |
+
import numpy as np
|
| 4 |
+
import pandas as pd
|
| 5 |
import streamlit as st
|
| 6 |
+
from sentence_transformers import SentenceTransformer
|
| 7 |
|
| 8 |
+
# st.set_page_config(layout="wide")
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
@st.cache_resource
|
| 12 |
+
def load_model():
|
| 13 |
+
return SentenceTransformer("TamedWicked/MathBERT_hr")
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
@st.cache_resource
|
| 17 |
+
def load_knowledge_base_df():
|
| 18 |
+
return pd.read_parquet("data/knowledge_base.parquet")
|
| 19 |
+
|
| 20 |
+
|
| 21 |
+
@st.cache_resource
|
| 22 |
+
def load_knowledge_base_index():
|
| 23 |
+
embeddings = joblib.load("data/knowledge_base_embeddings.pkl")
|
| 24 |
+
index = faiss.IndexFlatL2(embeddings.shape[1])
|
| 25 |
+
index.add(embeddings)
|
| 26 |
+
return index
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
def vector_search(query: list[str], model: SentenceTransformer, index, num_results=10):
|
| 30 |
+
vector = model.encode(list(query), show_progress_bar=False, convert_to_numpy=True)
|
| 31 |
+
D, I = index.search(np.array(vector).astype("float32"), k=num_results)
|
| 32 |
+
return D, I
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
def show_df_as_html(df: pd.DataFrame):
|
| 36 |
+
return df.to_html()
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
def show_df_as_markdown(df: pd.DataFrame):
|
| 40 |
+
return df.to_markdown()
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
model: SentenceTransformer = load_model()
|
| 44 |
+
df: pd.DataFrame = load_knowledge_base_df()
|
| 45 |
+
knowledge_index: np.array = load_knowledge_base_index()
|
| 46 |
+
|
| 47 |
+
query = st.text_input("Your math query:", value="Jesu li strukture koje su elementarno ekvivalentne izomorfne?")
|
| 48 |
+
if query:
|
| 49 |
+
D, I = vector_search([query], model, knowledge_index, num_results=5)
|
| 50 |
+
result = df[["Speech", "start_link"]].iloc[I[0]]
|
| 51 |
+
st.write(show_df_as_markdown(result), unsafe_allow_html=True)
|
data/knowledge_base.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1b60bdb181b8841e6d0d2e56f645c7070bf1017d056be4950f0facca9936be2a
|
| 3 |
+
size 999177
|
data/knowledge_base_embeddings.pkl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6bc7c3a5a91b52295c08bd22289a11d609f9a273e760c381041859bebe0c0223
|
| 3 |
+
size 58875105
|