Help defining the Config file for my own dataset inside Databricks

[strangan]

Dear RecBole Developers.

Can you help me with my configuration file? I'm trying to train a Sequential recommender on my own dataset, but I'm having trouble getting past the configuration step.

I'm using a Python Databricks notebook (so that i have access to GPU) and i've created a dataframe that looks like the one below --- WORKFLOWID, ACTIONID & SEQUENCE.

It's a pandas dataframe that i have written to an inter file in the dbfs filesystem that's tab delimited.

This is how i'm defining my config file:

paramter_dict = {
'model': 'Caser',
'dataset': 'nwc',
'data_path': "dbfs:/FileStore/recbole/nwc_action_sequences.inter",
'load_col': {'inter': ['WORKFLOWID', 'ACTIONID', 'SEQUENCE']},
'USER_ID_FIELD': 'WORKFLOWID',
'ITEM_ID_FIELD': 'ACTIONID',
'TIME_FIELD': 'SEQUENCE',
'MAX_ITEM_LIST_LENGTH': 3,
'neg_sample': 'None'
}

from logging import getLogger
from recbole.config import Config
from recbole.data import create_dataset
from recbole.data.utils import get_dataloader
from recbole.utils import init_logger, init_seed, get_model, get_trainer, set_color

config = Config(config_dict=paramter_dict)

But i get the following error when I run the above statement

ValueError: neg_sampling [{'uniform': 1}] should be None when the loss_type is CE.

Here's a screenshot of the entire cell

Can you point me in the right direction?

[Sherry-XLL]

@strangan Hello, you should set neg_sampling: ~ when the loss_type is CE.