Add Zipformer recipe for GigaSpeech #1254

yfyeung · 2023-09-14T14:22:22Z

Add Zipformer recipe for GigaSpeech
Upload checkpoints, exported models, logs, and bpe models trained using the GigaSpeech XL dataset to https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17
Update https://github.com/k2-fsa/icefall/blob/master/egs/gigaspeech/ASR/RESULTS.md
Update https://github.com/k2-fsa/icefall/blob/master/README.md
Update https://github.com/yfyeung/icefall/blob/phone2/egs/gigaspeech/ASR/README.md

desh2608 · 2023-10-02T14:28:47Z

@yfyeung do you have a pretrained model you could share?

yfyeung · 2023-10-17T12:15:59Z

@yfyeung do you have a pretrained model you could share?

@desh2608 Sure, check https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17

csukuangfj · 2023-10-18T01:39:11Z

README.md

@@ -148,8 +148,11 @@ in the decoding.

 ### GigaSpeech

-We provide two models for this recipe: [Conformer CTC model][GigaSpeech_conformer_ctc]
-and [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][GigaSpeech_pruned_transducer_stateless2].
+We provide three models for this recipe: [Zipformer]


Please add a link for [Zipformer].

OK, I removed it. It's redundant.

csukuangfj

Thanks!

The PR gives the state-of-the-art WER for gigapseech in icefall!

Left some minor comments.

csukuangfj · 2023-10-18T02:26:39Z

egs/gigaspeech/ASR/zipformer/asr_datamodule.py

@@ -0,0 +1,444 @@
+# Copyright      2021  Piotr Żelasko


Is it possible to replace it with a symlnik?

I have modified this file. In the past, GigaSpeech XL splits need to be merged. This one uses lhotse.mux.

csukuangfj · 2023-10-18T02:27:54Z

egs/gigaspeech/ASR/zipformer/export-onnx-ctc.py

@@ -0,0 +1,436 @@
+#!/usr/bin/env python3


Can you replace it with a symlink?

csukuangfj · 2023-10-18T02:28:11Z

egs/gigaspeech/ASR/zipformer/export-onnx-streaming.py

@@ -0,0 +1,775 @@
+#!/usr/bin/env python3


Can you replace it with a symlink?

csukuangfj · 2023-10-18T02:30:28Z

egs/gigaspeech/ASR/zipformer/export.py

@@ -0,0 +1,522 @@
+#!/usr/bin/env python3


Can you replace it with a symlink?

I changed some comments like huggingface link in it.

csukuangfj · 2023-10-18T02:30:54Z

egs/gigaspeech/ASR/zipformer/jit_pretrained.py

@@ -0,0 +1,280 @@
+#!/usr/bin/env python3
+# Copyright 2021-2023 Xiaomi Corporation (Author: Fangjun Kuang, Zengwei Yao)


Can you replace it with a symlink?

csukuangfj · 2023-10-18T02:32:12Z

egs/gigaspeech/ASR/zipformer/jit_pretrained_ctc.py

@@ -0,0 +1,436 @@
+#!/usr/bin/env python3


can you replace it with a symlink?

csukuangfj · 2023-10-18T02:32:29Z

egs/gigaspeech/ASR/zipformer/jit_pretrained_streaming.py

@@ -0,0 +1,273 @@
+#!/usr/bin/env python3


Can you replace it with a symlink?

csukuangfj · 2023-10-18T02:32:57Z

egs/gigaspeech/ASR/zipformer/onnx_check.py

@@ -0,0 +1,240 @@
+#!/usr/bin/env python3


Can you replace it with a symlink?

csukuangfj · 2023-10-18T02:34:36Z

egs/gigaspeech/ASR/RESULTS.md

+- normal-scaled model, number of model parameters: 65549011, i.e., 65.55 M
+
+You can find a pretrained model, training logs, decoding logs, and decoding results at:
+<https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17>


Could you upload the tensorboard log to
https://wandb.ai/site
and post a link to it here?

By the way, the link has not been posted. please leave a message when you think this PR is ready to merge.

https://wandb.ai/yifanyeung/icefall-asr-gigaspeech-zipformer-2023-10-20

OK, I have posted that.

csukuangfj · 2023-10-19T13:49:39Z

egs/gigaspeech/ASR/RESULTS.md

+- normal-scaled model, number of model parameters: 65549011, i.e., 65.55 M
+
+You can find a pretrained model, training logs, decoding logs, and decoding results at:
+<https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17>


Could you add a CI test for your model?

Ok, I have added.

Finish

yfyeung · 2023-10-20T15:07:48Z

Please re-review this PR. @csukuangfj

csukuangfj

Thanks!

yuyun2000 · 2024-06-12T03:56:38Z

I want a brand new streaming English model

yfyeung changed the title ~~Add GigaSpeech Zipformer~~ Add Zipformer recipe for GigaSpeech Sep 15, 2023

Yifan Yang added 2 commits October 17, 2023 17:58

Add Zipformer recipe for GigaSpeech

6f0f358

Finish

2732349

yfyeung force-pushed the phone2 branch from be71c5a to 2732349 Compare October 17, 2023 11:17

yfyeung requested a review from csukuangfj October 17, 2023 11:22

Yifan Yang and others added 6 commits October 17, 2023 19:32

update README

0c86ac9

Fix streaming_decode.py

00be865

Update README.md

327f475

Update RESULTS.md

9944b56

Update RESULTS.md

6eddab2

Fix for black

e71d008

csukuangfj reviewed Oct 18, 2023

View reviewed changes

csukuangfj previously requested changes Oct 18, 2023

View reviewed changes

yfyeung and others added 3 commits October 17, 2023 21:42

Merge branch 'k2-fsa:master' into phone2

e7414b9

Fix README.md

30a4dd2

Replace some with soft links

2ad1763

csukuangfj reviewed Oct 19, 2023

View reviewed changes

yfyeung and others added 4 commits October 20, 2023 08:39

Update RESULTS.md

7ad7e33

Merge branch 'k2-fsa:master' into phone2

91f4e52

Add ci

de8bb48

Delete .github/scripts/run-librispeech-zipformer-2023-10-17.sh

7de6e91

yfyeung requested a review from csukuangfj October 20, 2023 14:24

yfyeung added the zipformer For the upgraded zipformer recipes label Oct 20, 2023

Fix ci

cc42e4f

yfyeung added zipformer For the upgraded zipformer recipes and removed zipformer For the upgraded zipformer recipes labels Oct 20, 2023

csukuangfj approved these changes Oct 21, 2023

View reviewed changes

csukuangfj merged commit 416852e into k2-fsa:master Oct 21, 2023
36 checks passed

yfyeung deleted the phone2 branch October 21, 2023 12:46

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add Zipformer recipe for GigaSpeech #1254

Add Zipformer recipe for GigaSpeech #1254

yfyeung commented Sep 14, 2023 •

edited

Loading

desh2608 commented Oct 2, 2023

yfyeung commented Oct 17, 2023 •

edited

Loading

csukuangfj Oct 18, 2023 •

edited

Loading

yfyeung Oct 18, 2023

csukuangfj left a comment

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023 •

edited

Loading

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023

csukuangfj Oct 18, 2023

yfyeung Oct 18, 2023

csukuangfj Oct 19, 2023

yfyeung Oct 20, 2023

yfyeung Oct 20, 2023

csukuangfj Oct 19, 2023

yfyeung Oct 20, 2023

yfyeung commented Oct 20, 2023

csukuangfj left a comment

yuyun2000 commented Jun 12, 2024

		@@ -0,0 +1,280 @@
		#!/usr/bin/env python3
		# Copyright 2021-2023 Xiaomi Corporation (Author: Fangjun Kuang, Zengwei Yao)

Add Zipformer recipe for GigaSpeech #1254

Add Zipformer recipe for GigaSpeech #1254

Conversation

yfyeung commented Sep 14, 2023 • edited Loading

desh2608 commented Oct 2, 2023

yfyeung commented Oct 17, 2023 • edited Loading

csukuangfj Oct 18, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

csukuangfj left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

yfyeung Oct 18, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

yfyeung commented Oct 20, 2023

csukuangfj left a comment

Choose a reason for hiding this comment

yuyun2000 commented Jun 12, 2024

yfyeung commented Sep 14, 2023 •

edited

Loading

yfyeung commented Oct 17, 2023 •

edited

Loading

csukuangfj Oct 18, 2023 •

edited

Loading

yfyeung Oct 18, 2023 •

edited

Loading