MCPcopy Create free account
hub / github.com/THUDM/GLM / pad_seq

Method pad_seq

data_utils/datasets.py:563–567  ·  view source on GitHub ↗
(self, seq, pad_id=None)

Source from the content-addressed store, hash-verified

561 return tokens, targets, loss_masks, attention_mask
562
563 def pad_seq(self, seq, pad_id=None):
564 total_tokens = self.max_seq_len
565 num_pad_tokens = max(0, total_tokens - len(seq))
566 seq += [self.tokenizer.get_command('pad').Id if pad_id is None else pad_id] * (num_pad_tokens)
567 return seq
568
569
570class BlockDataset(data.Dataset):

Callers 1

__getitem__Method · 0.95

Calls 1

get_commandMethod · 0.80

Tested by

no test coverage detected