Quick Start: Knowledge-based Recommendation¶
For Knowledge-based Recommendation, we choose CKE model to show you how to train and test it on the ml-100k dataset from both API and source code.
Quick-start From API¶
1. Prepare your data:¶
Before running a model, firstly you need to prepare and load data. To help users quickly get start, RecBole has a build-in dataset ml-100k and you can directly use it. However, if you want to use other datasets, you can read Running New Dataset for more information.
Then, you need to set data config for data loading. You can create a yaml file called test.yaml and write the following settings:
# dataset config : Knowledge-based Recommendation
USER_ID_FIELD: user_id
ITEM_ID_FIELD: item_id
HEAD_ENTITY_ID_FIELD: head_id
TAIL_ENTITY_ID_FIELD: tail_id
RELATION_ID_FIELD: relation_id
ENTITY_ID_FIELD: entity_id
load_col:
inter: [user_id, item_id]
kg: [head_id, relation_id, tail_id]
link: [item_id, entity_id]
Knowledge-based recommendation models utilize KG information to make recommendations, so it needs to specify and load the kg information of the dataset.
You can get the kg triplets by calling dataset.kg_feat, it looks like:
head_id relation_id tail_id
0 m.04ctbw8 film.producer.film m.0bln8
1 m.0c3wmn film.film.actor m.02vxxgs
2 m.04t36 film.film_genre.films_in_this_genre m.05sbv3
3 m.08jl3y film.film.actor m.0v187kf
4 m.0513fcb film.film.actor m.0glmggf
... ... ... ...
91626 m.09v46zg film.film.prequel m.02862zk
91627 m.0jyx6 film.film.award_nomination m.0gr51
91628 m.043qq5y film.film.cinematography m.0bnth9_
91629 m.0b_zqd8 film.film.actor m.07xv9s
91630 m.0cr7n8 film.film.actor m.0g99qg5
RecBole also offer entity2id and id2entity, which map item_id and entity
2. Choose a model:¶
You can choose a model from our Model Introduction. Here we choose CKE model to demonstrate how to train and test the knowledge-based Recommendation,model.
Then, you need to set the parameter for CKE model. You can check the CKE and add the model settings into the test.yaml, like:
# model config
embedding_size: 64
kg_embedding_size: 64
reg_weights: [1e-02,1e-02]
If you want to run different models, you can read Running Different Models for more information.
3. Set training and evaluation config:¶
In RecBole, we support multiple training and evaluation methods. You can choose how to train and test model by simply setting the config.
Here we want to train and test the CKE model in training-validation-test method (optimize model parameters on the training set, do parameter selection according to the results on the validation set, and finally report the results on the test set) and evaluate the model performance by full ranking with all item candidates, so we can add the following settings into the test.yaml.
# Training and evaluation config
eval_args:
split: {'RS': [0.8, 0.1, 0.1]}
group_by: user
order: RO
mode: full
metrics: ['Recall', 'MRR', 'NDCG', 'Hit', 'Precision']
topk: 10
valid_metric: MRR@10
For more details of training and evaluation config, please refer to Training Settings and Evaluation Settings.
4. Run the model and collect the result¶
Now you have finished all the preparations, it’s time to run the model!
You can create a new python file (e.g., run.py), and write the following code:
from recbole.quick_start import run_recbole
run_recbole(model='CKE', dataset='ml-100k', config_file_list=['test.yaml'])
Then run the following command:
python run.py
And you will obtain the output like:
16 Jul 17:35 INFO ml-100k
The number of users: 944
Average actions of users: 106.04453870625663
The number of items: 1683
Average actions of items: 59.45303210463734
The number of inters: 100000
The sparsity of the dataset: 93.70575143257098%
Remain Fields: ['entity_id', 'user_id', 'item_id', 'head_id', 'relation_id', 'tail_id']
The number of entities: 34713
The number of relations: 26
The number of triples: 91631
The number of items that have been linked to KG: 1598
16 Jul 17:35 INFO [Training]: train_batch_size = [2048] negative sampling: [{'uniform': 1}]
16 Jul 17:35 INFO [Evaluation]: eval_batch_size = [4096] eval_args: [{'split': {'RS': [0.8, 0.1, 0.1]}, 'group_by': 'user', 'order': 'RO', 'mode': 'full'}]
16 Jul 17:35 INFO CKE(
(user_embedding): Embedding(944, 64)
(item_embedding): Embedding(1683, 64)
(entity_embedding): Embedding(34713, 64)
(relation_embedding): Embedding(26, 64)
(trans_w): Embedding(26, 4096)
(rec_loss): BPRLoss()
(kg_loss): BPRLoss()
(reg_loss): EmbLoss()
)
Trainable parameters: 2497920
Train 0: 100%|██████████████████████████████████████████████████| 40/40 [00:06<00:00, 5.73it/s]
16 Jul 17:36 INFO epoch 0 training [time: 6.98s, train_loss1: 27.7243, train_loss2: 21.9423, train_loss3: 0.0436]
Evaluate : 100%|██████████████████████████████████████████████| 472/472 [00:00<00:00, 3589.95it/s]
16 Jul 17:36 INFO epoch 0 evaluating [time: 0.13s, valid_score: 0.019500]
......
Train 86: 100%|██████████████████████████████████████████████████| 40/40 [00:07<00:00, 5.36it/s]
16 Jul 17:47 INFO epoch 86 training [time: 7.46s, train_loss1: 3.7211, train_loss2: 2.9693, train_loss3: 0.1157]
Evaluate : 100%|██████████████████████████████████████████████| 472/472 [00:00<00:00, 3165.16it/s]
16 Jul 17:47 INFO epoch 86 evaluating [time: 0.15s, valid_score: 0.376600]
16 Jul 17:47 INFO valid result:
recall@10 : 0.2083 mrr@10 : 0.3766 ndcg@10 : 0.2238 hit@10 : 0.7455 precision@10 : 0.1544
16 Jul 17:47 INFO Finished training, best eval result in epoch 75
16 Jul 17:47 INFO Loading model structure and parameters from saved\CKE-Jul-16-2022_17-35-57.pth
Evaluate : 100%|██████████████████████████████████████████████| 472/472 [00:00<00:00, 2936.85it/s]
16 Jul 17:47 INFO best valid : OrderedDict([('recall@10', 0.2115), ('mrr@10', 0.3832), ('ndcg@10', 0.2296), ('hit@10', 0.7391), ('precision@10', 0.1584)])
16 Jul 17:47 INFO test result: OrderedDict([('recall@10', 0.2483), ('mrr@10', 0.4895), ('ndcg@10', 0.2912), ('hit@10', 0.7709), ('precision@10', 0.1951)])
Finally you will get the model’s performance on the test set and the model file will be saved under the /saved. Besides, RecBole allows tracking and visualizing train loss and valid score with TensorBoard, please read the Use Tensorboard for more details.
The above is the whole process of running a model in RecBole, and you can read other docs for depth usage.
Quick-start From Source¶
Besides using API, you can also directly run the source code of RecBole. The whole process is similar to Quick-start From API. You can create a yaml file called test.yaml and set all the config as follow:
# dataset config : Knowledge-based Recommendation
USER_ID_FIELD: user_id
ITEM_ID_FIELD: item_id
HEAD_ENTITY_ID_FIELD: head_id
TAIL_ENTITY_ID_FIELD: tail_id
RELATION_ID_FIELD: relation_id
ENTITY_ID_FIELD: entity_id
load_col:
inter: [user_id, item_id]
kg: [head_id, relation_id, tail_id]
link: [item_id, entity_id]
# model config
embedding_size: 64
kg_embedding_size: 64
reg_weights: [1e-02,1e-02]
# Training and evaluation config
eval_args:
split: {'RS': [0.8, 0.1, 0.1]}
group_by: user
order: RO
mode: full
metrics: ['Recall', 'MRR', 'NDCG', 'Hit', 'Precision']
topk: 10
valid_metric: MRR@10
Then run the following command:
python run_recbole.py --model=CKE --dataset=ml-100k --config_files=test.yaml
And you will get the output of running the CKE model on the ml-100k dataset.
If you want to change the parameters, such as embedding_size, just
set the additional command parameters as you need:
python run_recbole.py --model=CKE --dataset=ml-100k --config_files=test.yaml --embedding_size=100