Add Mask to TextLastHiddenState #997
Merged
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
added 11 commits
November 7, 2022 10:09
…f code to fix this
Eugene-hu
reviewed
Jan 3, 2023
Eugene-hu
approved these changes
Jan 5, 2023
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Outdated suggestions cannot be applied.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.
Adds masking to the TextLastHiddenState Synapse.
For a return tensor of size
(batch_size, sequence_length, hidden_state), we can now optionally pass a mask as:bittensor.synapse.TextLastHiddenState(mask: Optional[List[int]]). The mask will apply across thesequence_lengthdimension, i.e. ifmask == [0], only the hidden state for the 0th element will be filled. Explicitly,return_tensor[:, 1:, :]would==0.Alternatively a list of non-consecutive integers can be passed, e.g.
bittensor.synapse.TextLastHiddenState(mask = [0, 5, 3])would only return non-empty tensors for sequence indices0,5, and3.This change can drastically decrease the network bandwidth.
Alternatively, the user can specify a random mask
d.text_last_hidden_state ( ... mask = random.choice( seq_len, k), ...)
This random masking can be used for adversarial resistance. Or Jepa style training.