alexlimh / slim Goto Github PK

Thanks for publishing the code of this important work.

I just follow the README to train the SLIM model on msmarco dataset again.
After generating the embeddings, I want to build an index on them but I don't no the details.
I try to find the instructions in pyserini, but the step about building index is omitted.
I really want to know how to build the index. I would appreciate your help.

I look forward to hearing from you.

Thank you for releasing the code of SLIM!
It is really a cool job.

I really want to train the SLIM model on msmarco dataset again, so I follow the README.
After creating the environment and splitting the training data into train and dev, I run the command provided by the README.
PYTHONPATH=.:$PYTHONPATH python dpr_scale/main.py -m \ --config-name msmarco_aws.yaml \ task=multiterm task/model=mtsplade_model \ task.model.sparse_mode=True \ task.in_batch_eval=True datamodule.num_test_negative=10 trainer.max_epochs=6 \ task.shared_model=True +task.cross_batch=False +task.in_batch=True \ +task.query_topk=20 +task.context_topk=20 \ +task.teacher_coef=0 +task.tau=1 \ +task.query_router_marg_load_loss_coef=0 +task.context_router_marg_load_loss_coef=0 \ +task.query_expert_load_loss_coef=1e-5 +task.context_expert_load_loss_coef=1e-5 \ datamodule.batch_size=8 datamodule.num_negative=7 \ trainer=gpu_1_host trainer.num_nodes=4 trainer.gpus=8

However, I encountered the following error :
Error executing job with overrides: ['task=multiterm', 'task/model=mtsplade_model', 'task.model.sparse_mode=True', 'task.in_batch_eval=True', 'datamodule.num_test_negative=10', 'trainer.max_epochs=6', 'task.shared_model=True', '+task.cross_batch=False', '+task.in_batch=True', '+task.query_topk=20', '+task.context_topk=20', '+task.teacher_coef=0', '+task.tau=1', '+task.query_router_marg_load_loss_coef=0', '+task.context_router_marg_load_loss_coef=0', '+task.query_expert_load_loss_coef=1e-05', '+task.context_expert_load_loss_coef=1e-05', 'datamodule.batch_size=8', 'datamodule.num_negative=7', 'trainer=gpu_1_host', 'trainer.num_nodes=4', 'trainer.gpus=8'] Error locating target 'dpr_scale.task.mtsplade_task.MultiTermRetrieverTask', set env var HYDRA_FULL_ERROR=1 to see chained exception. full_key: task Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.
Could you please tell me how to solve it ？
Thanks very much !!!