mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-18 21:44:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

update tensorboard and pre-models

Browse Source
This commit is contained in:
pkufool 2022-06-27 19:26:22 +08:00
parent 59b6be51b6
commit 2e5673f544
1 changed files with 17 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
egs/librispeech/ASR/RESULTS.md
View File

@ -1,6 +1,6 @@
## Results
### LibriSpeech BPE training results (Pruned Stateless streaming conformer RNN-T)
### LibriSpeech BPE training results (Pruned Stateless Streaming Conformer RNN-T)
#### [pruned_transducer_stateless](./pruned_transducer_stateless)
@ -8,6 +8,7 @@ See <https://github.com/k2-fsa/icefall/pull/380> for more details.
##### Training on full librispeech
The WERs are (the number in the table formatted as test-clean & test-other):
We only trained 25 epochs for saving time, if you want to get better results you can train more epochs.
| decoding method | left context | chunk size = 2 | chunk size = 4 | chunk size = 8 | chunk size = 16|
@ -37,7 +38,7 @@ The training command is:
--num-epochs 25
```
You can find the tensorboard log here <>
You can find the tensorboard log here <https://tensorboard.dev/experiment/ofxRakE6R7WHB1AoB8Bweg/>
The decoding command is:
```bash
@ -60,7 +61,7 @@ for chunk in 2 4 8 16; do
done
```
Pre-trained models, training and decoding logs, and decoding results are available at <>
Pre-trained models, training and decoding logs, and decoding results are available at <https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless_20220625>
#### [pruned_transducer_stateless2](./pruned_transducer_stateless2)
@ -68,6 +69,7 @@ See <https://github.com/k2-fsa/icefall/pull/380> for more details.
##### Training on full librispeech
The WERs are (the number in the table formatted as test-clean & test-other):
We only trained 25 epochs for saving time, if you want to get better results you can train more epochs.
| decoding method | left context | chunk size = 2 | chunk size = 4 | chunk size = 8 | chunk size = 16|
@ -97,7 +99,7 @@ The training command is:
--num-epochs 25
```
You can find the tensorboard log here <>
You can find the tensorboard log here <https://tensorboard.dev/experiment/hbltNS5TQ1Kiw0D1vcoakw/>
The decoding command is:
```bash
@ -120,13 +122,14 @@ for chunk in 2 4 8 16; do
done
```
Pre-trained models, training and decoding logs, and decoding results are available at <>
Pre-trained models, training and decoding logs, and decoding results are available at <https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless2_20220625>
#### [pruned_transducer_stateless3](./pruned_transducer_stateless3)
See <https://github.com/k2-fsa/icefall/pull/380> for more details.
##### Training on full librispeech (**Use giga_prob = 0.5)
##### Training on full librispeech (**Use giga_prob = 0.5**)
The WERs are (the number in the table formatted as test-clean & test-other):
| decoding method | left context | chunk size = 2 | chunk size = 4 | chunk size = 8 | chunk size = 16|
@ -159,7 +162,7 @@ The training command is (Note: this model was trained with mix-precision trainin
--giga-prob 0.5
```
You can find the tensorboard log here <>
You can find the tensorboard log here <https://tensorboard.dev/experiment/vL7dWVZqTYaSeoOED4rtow/>
The decoding command is:
```bash
@ -182,9 +185,10 @@ for chunk in 2 4 8 16; do
done
```
Pre-trained models, training and decoding logs, and decoding results are available at <>
Pre-trained models, training and decoding logs, and decoding results are available at <https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless3_giga_0.5_20220625>
##### Training on full librispeech (**Use giga_prob = 0.9**)
The WERs are (the number in the table formatted as test-clean & test-other):
| decoding method | left context | chunk size = 2 | chunk size = 4 | chunk size = 8 | chunk size = 16|
@ -216,7 +220,7 @@ The training command is:
--giga-prob 0.9
```
You can find the tensorboard log here <>
You can find the tensorboard log here <https://tensorboard.dev/experiment/WBGBDzt7SByRnvCBEfQpGQ/>
The decoding command is:
```bash
@ -239,7 +243,7 @@ for chunk in 2 4 8 16; do
done
```
Pre-trained models, training and decoding logs, and decoding results are available at <>
Pre-trained models, training and decoding logs, and decoding results are available at <https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless3_giga_0.9_20220625>
#### [pruned_transducer_stateless4](./pruned_transducer_stateless4)
@ -247,6 +251,7 @@ See <https://github.com/k2-fsa/icefall/pull/380> for more details.
##### Training on full librispeech
The WERs are (the number in the table formatted as test-clean & test-other):
We only trained 25 epochs for saving time, if you want to get better results you can train more epochs.
| decoding method | left context | chunk size = 2 | chunk size = 4 | chunk size = 8 | chunk size = 16|
@ -276,7 +281,7 @@ The training command is:
--num-epochs 25
```
You can find the tensorboard log here <>
You can find the tensorboard log here <https://tensorboard.dev/experiment/97VKXf80Ru61CnP2ALWZZg/>
The decoding command is:
```bash
@ -299,7 +304,7 @@ for chunk in 2 4 8 16; do
done
```
Pre-trained models, training and decoding logs, and decoding results are available at <>
Pre-trained models, training and decoding logs, and decoding results are available at <https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless4_20220625>
### LibriSpeech BPE training results (Pruned Stateless Conv-Emformer RNN-T)