ZhengkunTian/rnn-transducer

Issues

• Email error

  #27 opened a year ago by mflodzinski
  0

• 请问预训练和LM的代码在哪里啊？

  #25 opened 2 years ago by kartily
  0

• about loss

  #20 opened 4 years ago by cst781
  1

• 关于encoder代码的问题

  #24 opened 4 years ago by ZR-HH
  0

• 关于CER，在验证集上最好也就28？

  #23 opened 4 years ago by a122760
  0

• 请问训练过程中loss降到负了是正常的吗？

  #22 opened 4 years ago by zhoumumu
  1

• loss降低到 四五十后，不再下降了，在稳步上升。是啥情况。

  #21 opened 4 years ago by 786440445
  0

• Vocab

  #19 opened 4 years ago by Jiyongma06
  0

• Issue about eval.py

  #18 opened 4 years ago by Shengqiang-Li
  0

• 这要训练多少轮效果才好，感觉损失都好高，效果很差

  #16 opened 4 years ago by ZephyrChenzf
  8

• “RuntimeError: CUDA out of the memory. ” happens during training(epoch:0(34%~36%))

  #17 opened 4 years ago by jinsongpan
  0

• 一个batch内的识别结果相同是什么原因？

  #15 opened 4 years ago by ospeech
  7

• 请问160维MFCC特征是如何产生的？

  #8 opened 5 years ago by RenJinan
  3

• RNNT+pretrain+LM中的pretrain指的是预训练的什么呢？

  #14 opened 4 years ago by oshindow
  1

• LIne 140 in train.py

  #13 opened 5 years ago by oshindow
  1

• train.py return these errors

  #11 opened 5 years ago by oshindow
  5

• Is there a pre-trained model available?

  #12 opened 5 years ago by vasurobo
  1

• how to preprocess the dataset?

  #10 opened 5 years ago by w5688414
  2

• how to preprocess the dataset?

  #9 opened 5 years ago by w5688414
  0

• Structure of egs/timit/data/{train,dev,test}

  #7 opened 5 years ago by someshjaishwal
  4

• PRETRAINED MODEL

  #5 opened 5 years ago by tulasiram58827
  1

• RNNTLoss

  #6 opened 5 years ago by cdliang11
  1

• A question of recognize function implement

  #4 opened 5 years ago by huangnengCSU
  1

• loss值为负数

  #3 opened 5 years ago by Augus-smile
  3

• A question about vocab

  #2 opened 5 years ago by jackson1895
  1

• A question about beam search

  #1 opened 5 years ago by Marcovaldong
  2