gerardo
(Gerardo Garcia)
1
I’m not looking for advice on the competition.
I’m looking for technical advice.
The training includes 82 fields and 1 target
When I attempt to follow the tabular way
pytorch is looking to store 19 GB of space and my 1080ti includes only 11 GB
I don’t see any option to do batch size
Thanks!
1 Like
Ralph
2
Pass BS in the databunch
data = (TabularList.from_df(df, path=path, cat_names=cat_vars, cont_names=cont_vars, procs=procs)
.split_by_idx(valid_idx)
.label_from_df(cols=dep_var, label_cls=CategoryList)
.databunch(bs=128))
1 Like
gerardo
(Gerardo Garcia)
3
Well, according to the documentation
https://docs.fast.ai/basic_data.html#Factory-method
The default bs is 64
I guess I will have to try a lower number.