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

update multidataset zipformer results (#728)

Browse Source
This commit is contained in:
Weiji Zhuang 2022-12-02 17:40:42 +08:00 committed by GitHub
parent 6f71981667
commit 7700ddcb38
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 15 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
egs/librispeech/ASR/RESULTS.md
View File

@ -108,21 +108,25 @@ See <https://github.com/k2-fsa/icefall/pull/675> for more details.
[pruned_transducer_stateless8](./pruned_transducer_stateless8) [pruned_transducer_stateless8](./pruned_transducer_stateless8)
The tensorboard log can be found at The tensorboard log can be found at
<https://tensorboard.dev/experiment/y6kAPnN3S3OwvQxQqKQzsQ> <https://tensorboard.dev/experiment/3e9AfOcgRwOXpLQlZvHZrQ>
You can find a pretrained model, training logs, decoding logs, and decoding You can find a pretrained model, training logs, decoding logs, and decoding
results at: results at:
<https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless8-2022-11-14> <https://huggingface.co/WeijiZhuang/icefall-asr-librispeech-pruned-transducer-stateless8-2022-12-02>
You can use <https://github.com/k2-fsa/sherpa> to deploy it. You can use <https://github.com/k2-fsa/sherpa> to deploy it.
Number of model parameters: 70369391, i.e., 70.37 M Number of model parameters: 70369391, i.e., 70.37 M
| | test-clean | test-other | comment | | decoding method | test-clean | test-other | comment |
|----------------------|------------|-------------|----------------------------------------| |----------------------|------------|------------|--------------------|
| greedy search | 1.87 | 4.38 | --epoch 16 --avg 2 --max-duration 600 | | greedy_search | 1.81 | 4.18 | --epoch 20 --avg 4 |
| modified beam search | 1.81 | 4.34 | --epoch 16 --avg 2 --max-duration 600 | | fast_beam_search | 1.82 | 4.15 | --epoch 20 --avg 4 |
| fast beam search | 1.91 | 4.33 | --epoch 16 --avg 2 --max-duration 600 | | modified_beam_search | 1.78 | **4.08** | --epoch 20 --avg 4 |
| greedy_search | 1.84 | 4.3 | --epoch 19 --avg 8 |
| fast_beam_search |**1.77** | 4.25 | --epoch 19 --avg 8 |
| modified_beam_search | 1.81 | 4.16 | --epoch 19 --avg 8 |
The training commands are: The training commands are:
```bash ```bash
@ -142,9 +146,9 @@ export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"
The decoding commands are: The decoding commands are:
```bash ```bash
for m in greedy_search fast_beam_search modified_beam_search ; do for m in greedy_search fast_beam_search modified_beam_search; do
for epoch in 16; do for epoch in $(seq 20 -1 10); do
for avg in 2; do for avg in $(seq 9 -1 1); do
./pruned_transducer_stateless8/decode.py \ ./pruned_transducer_stateless8/decode.py \
--epoch $epoch \ --epoch $epoch \
--avg $avg \ --avg $avg \