recsys11.py

def recommend_item_for_user(model, user_id):
    m = model.eval().cpu()
    user_ids = torch.LongTensor([user2idx[u] for u in [user_id]*len(items)])
    item_ids = torch.LongTensor([item2idx[b] for b in items])
    remove = set(ratings[ratings[user_col] == user_id][item_col].values)
    preds = m(user_ids,item_ids).detach().numpy()
    pred_item = [(p,b) for p,b in sorted(zip(preds,items), reverse = True) if b not in remove]
    return pred_item

def recommend_user_for_item(model, item_id):
    m = model.eval().cpu()
    user_ids = torch.LongTensor([user2idx[u] for u in users])
    book_ids = torch.LongTensor([item2idx[b] for b in [item_id]*len(users)])
    remove = set(ratings[ratings[item_col] == book_id][user_col].values)
    preds = m(user_ids,item_ids).detach().numpy()
    pred_user = [(p,u) for p,u in sorted(zip(preds,users), reverse = True) if u not in remove]
    return pred_user