Upload 2 files

performer_pytorch/autoregressive_wrapper.py

@@ -4,8 +4,6 @@ from torch import nn
 import torch.nn.functional as F
 from torch.nn.utils.rnn import pad_sequence
 
-import pdb
-
 
 def exists(val):
     return val is not None
@@ -108,6 +106,4 @@ class AutoregressiveWrapper(nn.Module):
 
         loss = F.cross_entropy(out.transpose(1, 2), xo, ignore_index = self.ignore_index)
         
-        #pdb.set_trace()
-        
         return loss

performer_pytorch/performer_pytorch.py

@@ -12,7 +12,6 @@ from contextlib import contextmanager
 from local_attention import LocalAttention
 from performer_pytorch.reversible import ReversibleSequence, SequentialSequence
 
-import pdb
 
 try:
     from apex import amp
@@ -605,7 +604,6 @@ class PerformerLM(nn.Module):
         b, n, device = *x.shape, x.device
         assert n <= self.max_seq_len, f'sequence length {n} must be less than the max sequence length {self.max_seq_len}'
 
-        #pdb.set_trace()
         # token and positional embedding
         x = self.token_emb(x)
         if output_attentions: