Commits · Luigi/Streaming-Zipformer

add fp16 version of zipformer-zh-xlarge 06-30 and zipformer-zh 06-30

70a816c

Luigi commited on 16 days ago

do not make path check in choose_file() as model has not been downloaded yet

ac441da

Luigi commited on 16 days ago

add debug msg to choose_file()

6ffa7d7

Luigi commited on 16 days ago

fix for relative weight path

f95c465

Luigi commited on 16 days ago

1. Add two models

a6f5dad

Luigi commited on 16 days ago

add endpoint detection

548b7ed

Luigi commited on Jun 11

fallback to fp32 weights if int8 counterpart is not available

30a3b5d

Luigi commited on Jun 10

add csukuangfj/sherpa-onnx-streaming-zipformer-ar_en_id_ja_ru_th_vi_zh-2025-02-10">zipformer-ar_en_id_ja_ru_th_vi_zh model

ca1c0dd

Luigi commited on Jun 10

1. convert every hotword into zhcn for zhcn models

c335e79

Luigi commited on Jun 10

bugfix on hotword biasing

3ed4a9b

Luigi commited on Jun 9

update asr worker

e3fb82b

Luigi commited on Jun 9

update asr_worker.py

debc9c5

Luigi commited on Jun 9

add hotwords_file_path

1da82c9

Luigi commited on Jun 9

sanitize hotword list

6c8af7a

Luigi commited on Jun 9

remove \ from the end of line in hotword files

202cb59

Luigi commited on Jun 9

add hotwords (ie. speech context) demo

ab74fc2

Luigi commited on Jun 9

add bilingual small model (new default)

9fba0a6

Luigi commited on Jun 8

remove inexisting models

6296708

Luigi commited on Jun 8

remove final, as this demo perform asr forever

a1d489d

Luigi commited on Jun 8

fix model id

2632b6f

Luigi commited on Jun 8

add more models

454a10d

Luigi commited on Jun 8

fix cache directory access degnied issue

4a7594f

Luigi commited on Jun 8

fix local model path err

b26d94c

Luigi commited on Jun 8

get model from HF hub in runtime

ab9b679

Luigi commited on Jun 8

auto-detect mic sample rate

7c3f2af

Luigi commited on Jun 6

1. add to traditional chinese conversion 2. improve ui layout

231cd3a

Luigi commited on Jun 6

Revert "add simplified to traditional chinese conversion"

221a9c5

Luigi commited on Jun 6

add simplified to traditional chinese conversion

8ebc812

Luigi commited on Jun 6

resample mic audio to 16kHz

2a8a9a5

Luigi commited on Jun 6

commit app, Dockerfile and requirements.txt

2318eae

Luigi commited on Jun 6

Spaces:

Luigi
/

Streaming-Zipformer

Running

Commit History

add fp16 version of zipformer-zh-xlarge 06-30 and zipformer-zh 06-30

70a816c

do not make path check in choose_file() as model has not been downloaded yet

ac441da

add debug msg to choose_file()

6ffa7d7

fix for relative weight path

f95c465

1. Add two models

a6f5dad

add endpoint detection

548b7ed

fallback to fp32 weights if int8 counterpart is not available

30a3b5d

add csukuangfj/sherpa-onnx-streaming-zipformer-ar_en_id_ja_ru_th_vi_zh-2025-02-10">zipformer-ar_en_id_ja_ru_th_vi_zh model

ca1c0dd

1. convert every hotword into zhcn for zhcn models

c335e79

bugfix on hotword biasing

3ed4a9b

update asr worker

e3fb82b

update asr_worker.py

debc9c5

add hotwords_file_path

1da82c9

sanitize hotword list

6c8af7a

remove \ from the end of line in hotword files

202cb59

add hotwords (ie. speech context) demo

ab74fc2

add bilingual small model (new default)

9fba0a6

remove inexisting models

6296708

remove final, as this demo perform asr forever

a1d489d

fix model id

2632b6f

add more models

454a10d

fix cache directory access degnied issue

4a7594f

fix local model path err

b26d94c

get model from HF hub in runtime

ab9b679

auto-detect mic sample rate

7c3f2af

1. add to traditional chinese conversion 2. improve ui layout

231cd3a

Revert "add simplified to traditional chinese conversion"

221a9c5

add simplified to traditional chinese conversion

8ebc812

resample mic audio to 16kHz

2a8a9a5

commit app, Dockerfile and requirements.txt

2318eae

Commit History

add fp16 version of zipformer-zh-xlarge 06-30 and zipformer-zh 06-30 70a816c

do not make path check in choose_file() as model has not been downloaded yet ac441da

add debug msg to choose_file() 6ffa7d7

fix for relative weight path f95c465

1. Add two models a6f5dad

add endpoint detection 548b7ed

fallback to fp32 weights if int8 counterpart is not available 30a3b5d

add csukuangfj/sherpa-onnx-streaming-zipformer-ar_en_id_ja_ru_th_vi_zh-2025-02-10">zipformer-ar_en_id_ja_ru_th_vi_zh model ca1c0dd

1. convert every hotword into zhcn for zhcn models c335e79

bugfix on hotword biasing 3ed4a9b

update asr worker e3fb82b

update asr_worker.py debc9c5

add hotwords_file_path 1da82c9

sanitize hotword list 6c8af7a

remove \ from the end of line in hotword files 202cb59

add hotwords (ie. speech context) demo ab74fc2

add bilingual small model (new default) 9fba0a6

remove inexisting models 6296708

remove final, as this demo perform asr forever a1d489d

fix model id 2632b6f

add more models 454a10d

fix cache directory access degnied issue 4a7594f

fix local model path err b26d94c

get model from HF hub in runtime ab9b679

auto-detect mic sample rate 7c3f2af

1. add to traditional chinese conversion 2. improve ui layout 231cd3a

Revert "add simplified to traditional chinese conversion" 221a9c5

add simplified to traditional chinese conversion 8ebc812

resample mic audio to 16kHz 2a8a9a5

commit app, Dockerfile and requirements.txt 2318eae

add fp16 version of zipformer-zh-xlarge 06-30 and zipformer-zh 06-30

70a816c

do not make path check in choose_file() as model has not been downloaded yet

ac441da

add debug msg to choose_file()

6ffa7d7

fix for relative weight path

f95c465

1. Add two models

a6f5dad

add endpoint detection

548b7ed

fallback to fp32 weights if int8 counterpart is not available

30a3b5d

add csukuangfj/sherpa-onnx-streaming-zipformer-ar_en_id_ja_ru_th_vi_zh-2025-02-10">zipformer-ar_en_id_ja_ru_th_vi_zh model

ca1c0dd

1. convert every hotword into zhcn for zhcn models

c335e79

bugfix on hotword biasing

3ed4a9b

update asr worker

e3fb82b

update asr_worker.py

debc9c5

add hotwords_file_path

1da82c9

sanitize hotword list

6c8af7a

remove \ from the end of line in hotword files

202cb59

add hotwords (ie. speech context) demo

ab74fc2

add bilingual small model (new default)

9fba0a6

remove inexisting models

6296708

remove final, as this demo perform asr forever

a1d489d

fix model id

2632b6f

add more models

454a10d

fix cache directory access degnied issue

4a7594f

fix local model path err

b26d94c

get model from HF hub in runtime

ab9b679

auto-detect mic sample rate

7c3f2af

1. add to traditional chinese conversion 2. improve ui layout

231cd3a

Revert "add simplified to traditional chinese conversion"

221a9c5

add simplified to traditional chinese conversion

8ebc812

resample mic audio to 16kHz

2a8a9a5

commit app, Dockerfile and requirements.txt

2318eae