mirror of https://github.com/k2-fsa/icefall.git synced 2025-12-09 22:15:28 +00:00

History

Replace deprecated pytorch methods (#1814 )

* Replace deprecated pytorch methods

- torch.cuda.amp.GradScaler(...) => torch.amp.GradScaler("cuda", ...)
- torch.cuda.amp.autocast(...) => torch.amp.autocast("cuda", ...)

* Replace `with autocast(...)` with `with autocast("cuda", ...)`


Co-authored-by: Li Peng <lipeng@unisound.ai>

2024-12-16 10:24:16 +08:00

local

Traning with byte level BPE (TAL_CSASR) (#1033 )

2023-05-16 12:44:52 +08:00

pruned_transducer_stateless5

Replace deprecated pytorch methods (#1814 )

2024-12-16 10:24:16 +08:00

pruned_transducer_stateless7_bbpe

Replace deprecated pytorch methods (#1814 )

2024-12-16 10:24:16 +08:00

prepare.sh

typo fixed (#1334 )

2023-10-25 00:03:33 +08:00

README.md

[Ready to merge] Pruned transducer stateless5 recipe for tal_csasr (mix Chinese chars and English BPE) (#428 )

2022-06-28 11:02:10 +08:00

RESULTS.md

Traning with byte level BPE (TAL_CSASR) (#1033 )

2023-05-16 12:44:52 +08:00

shared

[Ready to merge] Pruned transducer stateless5 recipe for tal_csasr (mix Chinese chars and English BPE) (#428 )

2022-06-28 11:02:10 +08:00

README.md

Introduction

This recipe includes some different ASR models trained with TAL_CSASR.

./RESULTS.md contains the latest results.

Transducers

There are various folders containing the name transducer in this folder. The following table lists the differences among them.

	Encoder	Decoder	Comment
`pruned_transducer_stateless5`	Conformer(modified)	Embedding + Conv1d	same as pruned_transducer_stateless4 + more layers + random combiner

The decoder in transducer_stateless is modified from the paper Rnn-Transducer with Stateless Prediction Network. We place an additional Conv1d layer right after the input embedding layer.