mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-09 01:52:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'specaugmod_baseline' into randcombine1_expscale3_rework2c_maxabs1000_maxp0.95_noexp_convderiv2warmup_scale_0mean

Browse Source
This commit is contained in:
Daniel Povey 2022-03-15 23:47:03 +08:00
commit 1331199530
1 changed files with 6 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
egs/librispeech/ASR/transducer_stateless/diagnostics.py
View File

@ -135,16 +135,18 @@ def get_diagnostics_for_dim(dim: int, tensors: List[Tensor],
ans = [ '%.2g' % x for x in ans ] ans = [ '%.2g' % x for x in ans ]
ans = '[' + ' '.join(ans) + ']' ans = '[' + ' '.join(ans) + ']'
if stats_type == "value": if stats_type == "value":
norm = (stats ** 2).sum().sqrt().item()
mean_abs = stats.abs().mean().item()
# This norm is useful because it is strictly less than the largest # This norm is useful because it is strictly less than the largest
# sqrt(eigenvalue) of the variance, which we print out, and shows, # sqrt(eigenvalue) of the variance, which we print out, and shows,
# speaking in an approximate way, how much of that largest eigenvalue # speaking in an approximate way, how much of that largest eigenvalue
# can be attributed to the mean of the distribution. # can be attributed to the mean of the distribution.
ans += f', norm={norm:.2g}, mean_abs={mean_abs:.2g}' norm = (stats ** 2).sum().sqrt().item()
mean = stats.mean().item()
rms = (stats ** 2).mean().sqrt().item()
ans += f', norm={norm:.2g}, mean={mean:.2g}, rms={rms:.2g}'
else: else:
mean = stats.mean().item() mean = stats.mean().item()
ans += f', mean={mean:.2g}' rms = (stats ** 2).mean().sqrt().item()
ans += f', mean={mean:.2g}, rms={rms:.2g}'
return ans return ans