AttributeError: 'NoneType' object has no attribute 'repeat_interleave'

Beginners
1

My code -

import numpy as np
import torch
import faiss
from transformers import RagTokenizer, RagTokenForGeneration, DPRContextEncoder, DPRContextEncoderTokenizer
from datasets import Dataset

# Load DPR model and tokenizer
context_tokenizer = DPRContextEncoderTokenizer.from_pretrained("facebook/dpr-ctx_encoder-single-nq-base")
context_encoder = DPRContextEncoder.from_pretrained("facebook/dpr-ctx_encoder-single-nq-base")

# Load the FAISS index
index = faiss.read_index("faiss_index.index")

import os
os.environ['KMP_DUPLICATE_LIB_OK'] = 'TRUE'

# Define a function to retrieve documents using FAISS
def retrieve(query, index, k=5):
    # Convert query to embedding
    inputs = context_tokenizer(query, return_tensors="pt", padding=True, truncation=True, max_length=512)
    with torch.no_grad():
        query_embedding = context_encoder(**inputs).pooler_output.numpy().flatten()
    
    # Perform the search in FAISS index
    distances, indices = index.search(np.array([query_embedding]), k)
    return indices

# Load the RAG model and tokenizer
tokenizer = RagTokenizer.from_pretrained("facebook/rag-token-nq")
model = RagTokenForGeneration.from_pretrained("facebook/rag-token-nq")

# Function to query the RAG model with retrieved documents
def query_rag(query):
    # Retrieve document indices
    indices = retrieve(query, index)
    
    # Prepare the context for RAG model
    context = [chunks[i] for i in indices[0]]  # Use 'chunks' which should be globally accessible
    
    # Tokenize the context
    context_inputs = tokenizer(context, return_tensors="pt", padding=True, truncation=True, max_length=512, truncation_strategy="longest_first")
    
    # Prepare input_ids for the query
    input_ids = tokenizer(query, return_tensors="pt")["input_ids"]
    
    # Generate the response
    generated = model.generate(
        input_ids=input_ids,
        context_input_ids=context_inputs["input_ids"],
        context_attention_mask=context_inputs["attention_mask"],  # Include attention mask
        max_length=256
    )
    
    # Decode the generated response
    return tokenizer.batch_decode(generated, skip_special_tokens=True)[0]

# Example text file read and chunk creation
with open("rag_context.txt", "r") as f:
    context = f.read()

def chunk_text(text, chunk_size=500):
    words = text.split()
    for i in range(0, len(words), chunk_size):
        yield ' '.join(words[i:i + chunk_size])

chunks = list(chunk_text(context))

# Query the model
query = "Tell me about MPL's business model"
response = query_rag(query)
print(response)

In model.generate, I am getting AttributeError: 'NoneType' object has no attribute 'repeat_interleave'
Not able to understand what am I doing wrong here.

Code to generate fiass index

import numpy as np

with open("rag_context.txt", "r") as f:
    context = f.read()

def chunk_text(text, chunk_size=500):
    words = text.split()
    for i in range(0, len(words), chunk_size):
        yield ' '.join(words[i:i + chunk_size])

chunks = list(chunk_text(context))

import torch
from transformers import DPRContextEncoder, DPRContextEncoderTokenizer

# Load DPR model and tokenizer
context_tokenizer = DPRContextEncoderTokenizer.from_pretrained("facebook/dpr-ctx_encoder-single-nq-base")
context_encoder = DPRContextEncoder.from_pretrained("facebook/dpr-ctx_encoder-single-nq-base")

def get_embeddings(texts):
    embeddings = []
    for text in texts:
        inputs = context_tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=512)
        with torch.no_grad():
            emb = context_encoder(**inputs).pooler_output
        embeddings.append(emb.numpy().flatten())
    return embeddings

# Generate embeddings for each chunk
embedding_matrix = np.array(get_embeddings(chunks))

import faiss

# Create a FAISS index
dimension = embedding_matrix.shape[1]  # Dimensionality of embeddings
index = faiss.IndexFlatL2(dimension)
index.add(embedding_matrix)

# Save the FAISS index
faiss.write_index(index, "faiss_index.index")