File size: 3,658 Bytes
77d88ce 538e8da 77d88ce 538e8da 77d88ce | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 | import os, json
from typing import List
import numpy as np
import gradio as gr
import faiss
from sentence_transformers import SentenceTransformer
HF_TOKEN = os.getenv("HF_TOKEN")
# # ---------- Paths ----------
# APP_DIR = os.path.dirname(__file__)
# CORPUS_JSON = os.path.join(APP_DIR, "corpus.json")
# FAISS_MAIN = os.path.join(APP_DIR, "faiss_ip_768.index")
# # ---------- Load corpus ----------
# with open(CORPUS_JSON, "r", encoding="utf-8") as f:
# corpus = json.load(f) # [{"title", "text"}, ...]
# # ---------- Load FAISS index ----------
# if not os.path.exists(FAISS_MAIN):
# raise FileNotFoundError(f"Missing FAISS index at {FAISS_MAIN}")
# index = faiss.read_index(FAISS_MAIN)
# # Infer dimension from index
# EMB_DIM = index.d
# ---------- Model ----------
model = SentenceTransformer("google/embeddinggemma-300m", token=HF_TOKEN)
# ---------- 1. Display Similarity ----------
def do_similarity(text_a: str, text_b: str) -> float:
embeddings = model.encode(
[text_a, text_b],
normalize_embeddings=True,
convert_to_numpy=True
)
# The dot product of two normalized vectors is their cosine similarity.
similarity_score = np.dot(embeddings[0], embeddings[1])
return float(similarity_score)
# # ---------- 2. Semantic Search ----------
# def do_search(query: str, top_k: int = 5) -> List[List[str]]:
# if not query.strip():
# return []
# ##################################
# # INSERT YOUR CODE HERE
# ##################################
# rows = []
# for score, i in zip(scores[0], idxs[0]):
# if i == -1: continue
# item = corpus[i]
# snippet = item["text"][:380] + ("…" if len(item["text"]) > 380 else "")
# rows.append([f"{score:.4f}", item["title"], snippet])
# return rows
# ---------- UI ----------
with gr.Blocks(theme=gr.themes.Soft(), title="Embeddings Lab") as demo:
with gr.Row():
with gr.Column(scale=3):
gr.Markdown("""
# 🧬 Google Gemma Embeddings Lab
Compare the semantic similarity between two texts using Google Gemma embeddings.
""")
with gr.Column(scale=1):
gr.Markdown("""
### ℹ️ Instructions
- Enter two texts to compare their similarity.
- Click **Compute Similarity** to see the cosine similarity score.
- Example: Try comparing a question and its answer!
""")
with gr.Tabs():
with gr.TabItem("Display Similarity"):
with gr.Group():
a = gr.Textbox(lines=4, label="Text A", placeholder="Enter first text...", value="What is AI?", show_copy_button=True)
b = gr.Textbox(lines=4, label="Text B", placeholder="Enter second text...", value="Artificial Intelligence is the simulation of human intelligence in machines.", show_copy_button=True)
sim_btn = gr.Button("🔍 Compute Similarity", variant="primary")
sim_out = gr.Number(label="Cosine similarity", precision=4)
sim_btn.click(lambda x, y: do_similarity(x, y), [a, b], sim_out, api_name="similarity")
# with gr.TabItem("Semantic Search"):
# q = gr.Textbox(label="Query")
# topk = gr.Slider(1, 20, value=5, step=1, label="Top-K")
# run = gr.Button("Search")
# out = gr.Dataframe(headers=["score", "title", "snippet"], wrap=True)
# run.click(lambda query, k: do_search(query, int(k)), [q, topk], out, api_name="search")
if __name__ == "__main__":
demo.launch(server_name="0.0.0.0", server_port=7860)
|