mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-19 05:54:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add temperature in doc

Browse Source
This commit is contained in:
Erwan 2022-07-11 15:13:46 +02:00
parent fd4deeab95
commit e5cadba145
1 changed files with 12 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
egs/librispeech/ASR/pruned_transducer_stateless2/beam_search.py
View File

@ -126,6 +126,8 @@ def fast_beam_search_nbest_LG(
use_double_scores: use_double_scores:
True to use double precision for computation. False to use True to use double precision for computation. False to use
single precision. single precision.
temperature:
Softmax temperature.
Returns: Returns:
Return the decoded result. Return the decoded result.
""" """
@ -244,6 +246,8 @@ def fast_beam_search_nbest(
use_double_scores: use_double_scores:
True to use double precision for computation. False to use True to use double precision for computation. False to use
single precision. single precision.
temperature:
Softmax temperature.
Returns: Returns:
Return the decoded result. Return the decoded result.
""" """
@ -331,7 +335,8 @@ def fast_beam_search_nbest_oracle(
nbest_scale: nbest_scale:
It's the scale applied to the lattice.scores. A smaller value It's the scale applied to the lattice.scores. A smaller value
yields more unique paths. yields more unique paths.
temperature:
Softmax temperature.
Returns: Returns:
Return the decoded result. Return the decoded result.
""" """
@ -404,6 +409,8 @@ def fast_beam_search(
Max states per stream per frame. Max states per stream per frame.
max_contexts: max_contexts:
Max contexts pre stream per frame. Max contexts pre stream per frame.
temperature:
Softmax temperature.
Returns: Returns:
Return an FsaVec with axes [utt][state][arc] containing the decoded Return an FsaVec with axes [utt][state][arc] containing the decoded
lattice. Note: When the input graph is a TrivialGraph, the returned lattice. Note: When the input graph is a TrivialGraph, the returned
@ -809,6 +816,8 @@ def modified_beam_search(
encoder_out before padding. encoder_out before padding.
beam: beam:
Number of active paths during the beam search. Number of active paths during the beam search.
temperature:
Softmax temperature.
Returns: Returns:
Return a list-of-list of token IDs. ans[i] is the decoding results Return a list-of-list of token IDs. ans[i] is the decoding results
for the i-th utterance. for the i-th utterance.
@ -1072,6 +1081,8 @@ def beam_search(
A tensor of shape (N, T, C) from the encoder. Support only N==1 for now. A tensor of shape (N, T, C) from the encoder. Support only N==1 for now.
beam: beam:
Beam size. Beam size.
temperature:
Softmax temperature.
Returns: Returns:
Return the decoded result. Return the decoded result.
""" """