how can I use the self-made factor with SVD in python

I am trying to make a recommendation system by SVD.

I made a concatenated dataframe of LDA vector from items descriptions and image features of items.

Is there any possible I can use the dataframe I made as initial embedding factor?

I've tried to google key word like "svd embedding", but I cannot find such information or code.

from surprise import SVD
from surprise import Dataset, Reader
from surprise.model_selection import cross_validate, train_test_split
from tqdm import tqdm
reader = Reader(rating_scale=(1,5))
data = Dataset.load_from_df(trans_rating, reader)
trainset, testset = train_test_split(data, test_size=.2)
model = SVD(n_factors=100)
model.fit(trainset)
for user in tqdm(user_to_row_idx['user']):
   for item in item_to_row_idx['item']:
       rating = model.predict(user,item)
       row=[user,item,rating.est]
       df=pd.DataFrame([row],columns=['user','item','rating'])
       recommendation=recommendation.append(df,ignore_index = True)

And this is my dataset. enter image description here