-
你好,很期待你的joint ctc-attention预训练模型的开源,请问下,开源模型最近能看得到吗
-
# speech recognition
- Soltau, Hagen, Hank Liao, and Hasim Sak. "Neural Speech Recognizer: Acoustic-to-Word LSTM Model for Large Vocabulary Speech Recognition." arXiv preprint arXiv:1610.09975 (201…
-
Hi,
I'm working with the streaming Conformer models based on the Aishell1 ASR1 recipes(https://github.com/espnet/espnet/blob/master/egs2/aishell/asr1/conf/train_asr_streaming_conformer.yaml) but mo…
-
请问这个项目可以直接训练吗,训练效果怎么样。
-
- [x] Seeding (Zhaofeng)
- [x] `Dataset`s should probably share `char2idx` as we want unseen chars to be generated as well (Joseph)
- [x] Sort through the data points in increaseing video length
- …
-
Hi,
When I use streamming transformer to train the model,
unit: word
when I decode the result, using streaming_score.sh. I got an error:
```
Original utterance num: 2000
Removed 0 empty utteran…
-
Once my ASR model is trained, I run the `espnet2/bin/asr_inference.py` script for decoding. However, I am a bit confused about these warning that I getting quite oftenly:
```WARNING:root:decoding m…
-
I created a confomer-rnn-trasducer recipe using CSJ and run the training process.
When I evaluated the created model, the CERs for eval1, eval2, and eval3 were higher than the [original paper](https:…
-
I am getting a lot of deletions in my RNN-T training/decoding setup relative to Transformer/Conformer. The data is the "Malach" corpus; about 200 hours of English but accented speech from Holocaust su…
-
Thank you for ESPnet team's continuous support. I have been using ESPnet and ESPnet2 for silent speech recogntion tasks for two years. (https://github.com/espnet/espnet/issues/1926)
I did visual sp…