mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-12-11 06:55:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

Bug fix in caching_eval, may make no difference.

Browse Source
This commit is contained in:
Daniel Povey 2022-12-15 23:32:29 +08:00
parent f5d4fb092d
commit 6caaa4e9c6
1 changed files with 0 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
egs/librispeech/ASR/pruned_transducer_stateless7/scaling.py
View File

@ -245,7 +245,6 @@ class CachingEvalFunction(torch.autograd.Function):
# Caution: this assumes you are not going to use any random numbers from torch (for any purpose # Caution: this assumes you are not going to use any random numbers from torch (for any purpose
# that matters in the forward pass), e.g. there should be no dropout. # that matters in the forward pass), e.g. there should be no dropout.
ctx.random_state = random.getstate() ctx.random_state = random.getstate()
ctx.save_for_backward(x)
# we are inside torch.no_grad() here, so the following won't create the computation graph. # we are inside torch.no_grad() here, so the following won't create the computation graph.
y = m(x) y = m(x)
ctx.save_for_backward(x, y) ctx.save_for_backward(x, y)