Libri100 recipe for standalone Transducer

[b-flo]

Add scripts + configs for streaming and offline Transducer. The second model is training, I'll add results to the README and update both models to hf.
Also, I added some minor fixes to the PR, it won't impact previous versions in any way.

I made a separate directory for asr_transducer1 here because we need some separation between the two versions of Transducer in ESPnet2 but I don't really have an opinion on the design. It makes sense because we defined this version as a new "task" but we could also keep asr1 and use some prefixes for the standalone version.

[codecov]

Codecov Report

Merging #4698 (ce88312) into master (3297e10) will increase coverage by 0.63%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #4698      +/-   ##
==========================================
+ Coverage   74.36%   74.99%   +0.63%     
==========================================
  Files         654      655       +1     
  Lines       58347    58546     +199     
==========================================
+ Hits        43391    43909     +518     
+ Misses      14956    14637     -319     

Flag	Coverage Δ
test_integration_espnet1	66.24% <ø> (-0.05%)	⬇️
test_integration_espnet2	47.65% <ø> (+0.70%)	⬆️
test_python	65.28% <ø> (+0.09%)	⬆️
test_utils	23.27% <ø> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

see 15 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[b-flo]

@ftshijt @sw005320 If this design is okay, the PR can be merged. Otherwise I'll make the requested changes.

I'll update HF links in another PR, I'm currently running some additional experiments.

[b-flo]

BPE size is 500 vs 5000 for the CTC-Att baseline model. I guess we can further improve results with bigger BPE size but I don't have enough resource for that.

[sw005320]

@ftshijt @sw005320 If this design is okay, the PR can be merged. Otherwise I'll make the requested changes.

I'll update HF links in another PR, I'm currently running some additional experiments.

Sorry, I'll have some reviews after the ICASSP deadline. Can you wait for a week or so?

[mergify]

This pull request is now in conflict :(

[b-flo]

It's more stable on my side so I'll start training different model architectures and settings before opening the corresponding PR. I'll use this branch (and message) to keep track of the best models.

Model	Mode	Num. params.	BPE	dev_clean (%WER)	dev_other (%WER)	test_clean (%WER)	test_other (%WER)
E-Branchformer/Transformer CTC-Att	offline	38.47M	5000	6.1	16.7	6.3	17.0
Conformer/RNN (old)	offline	30.56M	500	5.9	17.6	6.4	17.9
Conformer/RNN (new)	offline	30.53M	500	5.8	16.9	6.0	17.0
E-Branchformer/RNN (new)	offline	29.12M	500	5.7	16.8	6.0	17.1
E-Branchformer/MEGA (tmp)	offline	40.66M	500	5.7	16.5	6.0	16.7

[b-flo]

@sw005320 If we are okay with the design (i.e.: adding a asr_transducer1 for standalone Transducer recipes), I think the PR can be merged.

Pre-trained model links will be added after I finish experiments with RWKV and I need to re-train the models anyways. At least, users have some examples for the standalone version and Librispeech-100 for now (which was asked before).