Upload 6 files

Initial commit: Added Luxury Decor Assistant RAG app with FAISS + DeBERTa + LLaMA-2

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+deberta_faiss.index filter=lfs diff=lfs merge=lfs -text
+Online[[:space:]]Retail.xlsx filter=lfs diff=lfs merge=lfs -text

Online Retail.xlsx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22d7d032ef5e8e9d039c16db624fc570ec2bacaf42e5dd3055cb5732971de78a
+size 2220651

app.py

@@ -0,0 +1,67 @@
+
+import gradio as gr
+import numpy as np
+import pandas as pd
+import faiss
+import torch
+from transformers import AutoTokenizer, AutoModel, pipeline
+
+# Load FAISS index and product data
+index = faiss.read_index("deberta_faiss.index")
+text_data = pd.read_csv("deberta_text_data.csv")["Retrieved Text"].tolist()
+
+# Load DeBERTa model and tokenizer for embedding
+deberta_model_name = "microsoft/deberta-v3-base"
+deberta_tokenizer = AutoTokenizer.from_pretrained(deberta_model_name)
+deberta_model = AutoModel.from_pretrained(deberta_model_name).to("cpu")
+
+# Load LLaMA-2 tokenizer and pipeline
+llama_model_name = "meta-llama/Llama-2-7b-chat-hf"
+llama_tokenizer = AutoTokenizer.from_pretrained(llama_model_name)
+llama_pipeline = pipeline("text-generation", model=llama_model_name, tokenizer=llama_tokenizer, device=-1)
+
+# Function to generate embeddings from DeBERTa
+def generate_embeddings(queries):
+    tokens = deberta_tokenizer(queries, return_tensors="pt", padding=True, truncation=True).to("cpu")
+    with torch.no_grad():
+        outputs = deberta_model(**tokens).last_hidden_state.mean(dim=1).cpu().numpy().astype("float32")
+    return outputs
+
+# Define the RAG response logic
+def generate_response(user_query):
+    query_embedding = generate_embeddings([user_query])
+    faiss.normalize_L2(query_embedding)
+    distances, indices = index.search(query_embedding, k=5)
+    retrieved_docs = [text_data[idx] for idx in indices[0]]
+    context = ", ".join(set(retrieved_docs))
+
+    prompt = f"""
+    Using the following product descriptions:
+    {context}
+
+    Carefully craft a well-structured response to the following question:
+
+    **Question:** {user_query}
+
+    **Instructions:**
+    1. Incorporate **all** retrieved product descriptions.
+    2. Use a **formal yet engaging** tone.
+    3. Provide **practical & creative** luxury decor ideas.
+    4. Ensure a **cohesive & detailed response.**
+
+    **Your response:**
+    """
+
+    result = llama_pipeline(prompt, max_length=512, truncation=True, do_sample=True)[0]["generated_text"]
+    return result
+
+# Gradio UI
+interface = gr.Interface(
+    fn=generate_response,
+    inputs=gr.Textbox(lines=2, placeholder="Ask a question about luxury home decor..."),
+    outputs="text",
+    title="Luxury Decor Assistant (RAG)",
+    description="Ask your luxury decor questions based on real product descriptions!"
+)
+
+interface.launch()

deberta_embeddings.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bd48c0b806882c35bacac456c8885e228e2daf5dedf3bb6ee5a29a705fd83ec
+size 87496832

deberta_faiss.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0961f49569103d709230403f8ca053da05fc5d73aebdf8f389a6590c7b01a29
+size 87496749

requirements.txt

@@ -0,0 +1,6 @@
+gradio
+faiss-cpu
+torch
+transformers
+pandas
+numpy