Feature/8mi trainer #834
Open
Feature/8mi trainer #834
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dpressel
force-pushed
the
feature/8mi-trainer
branch
2 times, most recently
from
c70b0b2
to
a658ea8
Compare
dpressel
force-pushed
the
feature/8mi-trainer
branch
from
3101a57
to
b2b31f9
Compare
WIP WIP: remap tagger and classifier onto Trainer oops fix issue with span F1 aggregation in Trainer adds support for other loss functions like KLDiv this is useful for cases like distillation where we can have soft targets. pass kwargs into target use forward function option whether to rm wrapper support overriding the train target This should fix multiworker mismatch on reload feelgood types fix first batch accum allow no early stopping global_step fix, clean examples, factor up more cleanup fix includes in addon rm dist code outside 8mi trainer, WIP dataset use native loaders via mead pseudo fix for showing examples fix default and backend arg in paired reader bye six + tmp working non-native LM loader add backend option LM is TF native fix test remove and simplify tf trainers and fix trim issue be a little tricky with TF native we cant switch it on with TF 1.x . explore more modular refactoring WIP WIP: remap tagger and classifier onto Trainer oops fix issue with span F1 aggregation in Trainer adds support for other loss functions like KLDiv this is useful for cases like distillation where we can have soft targets. pass kwargs into target use forward function option whether to rm wrapper support overriding the train target This should fix multiworker mismatch on reload feelgood types fix first batch accum allow no early stopping global_step fix, clean examples, factor up more cleanup fix includes in addon rm dist code outside 8mi trainer, WIP dataset use native loaders via mead pseudo fix for showing examples fix default and backend arg in paired reader bye six + tmp working non-native LM loader add backend option LM is TF native fix test remove and simplify tf trainers and fix trim issue .
dpressel
force-pushed
the
feature/8mi-trainer
branch
from
b2b31f9
to
0432595
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
mead3_pytorch, supports future (distributed) training using 8mi viafit_func=mead3sixneg_log_loss