issues
search
k2-fsa
/
snowfall
Moved to https://github.com/k2-fsa/icefall
Apache License 2.0
143
stars
42
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
n-best rescore with transformer lm
#201
glynpu
opened
3 years ago
3
Topology problem?
#200
danpovey
closed
3 years ago
2
Support using k2.intersect_dense_pruned in computing LF-MMI loss.
#199
csukuangfj
closed
3 years ago
9
[WIP] Implement 2nd pass training using 1-best decoding results from the 1st pass network
#198
csukuangfj
opened
3 years ago
0
Concatenate valid cuts, reduce chance of OOM
#197
danpovey
closed
3 years ago
5
Fix interaction of alignment model with accum_grad
#196
danpovey
closed
3 years ago
0
Topology choices
#195
danpovey
opened
3 years ago
0
Inconsistency between the reported loss and the actual loss used for gradient computation
#194
pzelasko
opened
3 years ago
2
[WIP] Add iterated loss
#193
zhu-han
opened
3 years ago
0
Fix hardcoded librispeech lang dir in local/prepare_lang.sh
#192
pzelasko
closed
3 years ago
0
Building lexicons in Python
#191
pzelasko
opened
3 years ago
16
Fix save_checkpoint()
#190
csukuangfj
closed
3 years ago
0
Rework code that deals with conformer positional embedding for clarity and efficiency
#189
danpovey
closed
3 years ago
1
Assign a default value to scaler in save_checkpoint().
#188
csukuangfj
closed
3 years ago
0
error when run "python3 ./mmi_bigram_train.py"
#187
yujunlhz
closed
3 years ago
4
error when PytorchStreamReader failed reading file
#186
yujunlhz
opened
3 years ago
2
Have a look at SpeechBrain
#185
csukuangfj
opened
3 years ago
1
Fix transformer/conformer bugs when normalize_before is false
#184
zhu-han
closed
3 years ago
0
Why do ctc.py and mmi.py use use_double_scores=True?
#183
galv
opened
3 years ago
5
Add VGG frontend
#182
zhu-han
closed
3 years ago
4
Vgg frontend
#181
danpovey
closed
3 years ago
5
torch.multiprocessing.spawn and changing files
#180
danpovey
opened
3 years ago
0
Label smoothing for LF-MMI
#179
danpovey
opened
3 years ago
14
Fix output_beam_size in decoding.
#178
csukuangfj
closed
3 years ago
0
Fix att_rate.
#177
csukuangfj
closed
3 years ago
1
Error when att-rate > 0
#176
zhichaowang
closed
3 years ago
10
Fix grad scaler merge issues
#175
pzelasko
closed
3 years ago
0
Use pre-trained CTC model for conformer training.
#174
csukuangfj
closed
3 years ago
5
Support ContextNet
#173
zhu-han
closed
3 years ago
8
Improve training speed by pre-computing `compose(ctc_topo, P, L)`
#172
csukuangfj
closed
3 years ago
6
Mixed precision training
#171
pzelasko
closed
3 years ago
4
Convergence & decoding problem
#170
zhichaowang
opened
3 years ago
16
Conformer & convergence
#169
danpovey
opened
3 years ago
0
Install Error
#168
zhichaowang
closed
3 years ago
3
Fix loading checkpoints with DDP
#167
pzelasko
closed
3 years ago
0
Invoke k2.intersect_dense and get_tot_scores only once in LFMMILoss
#166
csukuangfj
closed
3 years ago
8
Speeding up training
#165
danpovey
opened
3 years ago
0
Print default values in the help message.
#164
csukuangfj
closed
3 years ago
1
Print default args
#163
danpovey
closed
3 years ago
0
Collect validation objf from all nodes in DDP training.
#162
csukuangfj
closed
3 years ago
3
Initialization of last layer to zero
#161
danpovey
opened
3 years ago
4
how to convert LG.aux_labels of type tensor to _k2.RaggedInt
#160
Jarvan-Wang
closed
3 years ago
7
The LG.arcs_as_tensor().shape becomes extreme small after compose with ctc_topo
#159
Jarvan-Wang
closed
3 years ago
8
Make DDP master port configurable
#158
danpovey
closed
3 years ago
9
Revert "Make master port configurable"
#157
danpovey
closed
3 years ago
1
Make master port configurable
#156
danpovey
closed
3 years ago
0
Merge changes related to ddp from pre_refactor
#155
danpovey
closed
3 years ago
5
Full-libri default true?
#154
danpovey
opened
3 years ago
9
Sampler type mismatch
#153
danpovey
closed
3 years ago
1
DDP address in use
#152
danpovey
opened
3 years ago
20
Previous
Next