Sometimes I just want to call a pipeline with out the extra stuff eval_, train_ or infer_ apply do (iterating over the batch, disabling gradient etc).
This is especially useful when combining pipelines with non-padl code, without that, it's a bit awkward to do.
wuhu
commented
2 years ago
🛰️ Feature
Sometimes I just want to call a pipeline with out the extra stuff
eval_,train_orinfer_apply do (iterating over the batch, disabling gradient etc). This is especially useful when combining pipelines with non-padl code, without that, it's a bit awkward to do.