smeetrs/deep_avsr

Issues

• audio-visual.pt file

  #55 opened 4 months ago by zlwxwxsdzrx
  4

• About loss curve

  #53 opened 5 months ago by rwbybxx
  4

• How AO and VO model are trained?

  #54 opened 5 months ago by omerfdeniz
  1

• RuntimeError: cublas runtime error

  #52 opened 5 months ago by crucian-star
  2

• Error happening when evaluating

  #51 opened 7 months ago by wcycqjy
  1

• preprocess error

  #50 opened a year ago by ouwen18
  2

• When pretrain VO model, loss reduce a few epochs, but immediately increase and remain.

  #43 opened 2 years ago by FL990
  3

• Questions about meanings of variable in beam search

  #49 opened a year ago by HaoDot
  0

• Where is the seq2seq model at 'audi_visual' or 'video only'?

  #48 opened a year ago by HaoDot
  1

• The python version 3.6.9 is not supported anymore.

  #47 opened 2 years ago by vishwa3344
  2

• Why set "AUDIO_ONLY_PROBABILITY" and "VIDEOO_ONLY_PROBABILITY" these two parameters?

  #46 opened 2 years ago by gmh8000
  2

• CUDA out of memory on preprocess.py

  #45 opened 2 years ago by InterStella0
  1

• What does <EOS> and " " mean?

  #44 opened 2 years ago by FL990
  1

• Train and validation WER both remain 1 while training VO model

  #6 opened 4 years ago by smeetrs
  5

• Language Model Execution Never Ends

  #40 opened 3 years ago by padmanabhankrishnamurthy
  8

• The Validation WER in pretrain is always 1.0

  #42 opened 3 years ago by lightning830
  5

• Question about the implement

  #41 opened 3 years ago by zero0kiriyu
  1

• Transferring to train set was not going well

  #37 opened 3 years ago by amadeusuzx
  9

• Details of visual fronted model

  #39 opened 3 years ago by gmh8000
  2

• ctc_loss problem when running on multiple GPUs

  #38 opened 3 years ago by gmh8000
  3

• Transformer Decoder in jointDecoder in avnet

  #36 opened 3 years ago by lastlap
  1

• WER after pretraining and training

  #30 opened 3 years ago by lastlap
  7

• The performance when doing curriculum learning

  #28 opened 3 years ago by xichenpan
  6

• When I train the ao model ,the WER and CER are always 1.

  #15 opened 4 years ago by liuzeyang0311
  3

• the result under noisy audio

  #27 opened 3 years ago by xjwla
  1

• Selection of videos

  #26 opened 3 years ago by zhanghousen
  2

• question

  #23 opened 4 years ago by iflyBird
  1

• The error when use preprocess.py

  #19 opened 4 years ago by xjwla
  3

• The error when use preprocess.py

  #20 opened 4 years ago by xjwla
  0

• Out of Memory errors during pretraining

  #3 opened 4 years ago by kasri-mids
  19

• How to initialize encoders with AO and VO sets

  #18 opened 4 years ago by arunm95
  1

• preprocess for lrs2 dataset

  #16 opened 4 years ago by javadpeyman
  2

• Pretrained models not available

  #8 opened 4 years ago by ashutosh1919
  7

• When I was training the ao model,the WER and CER were always 1.000

  #14 opened 4 years ago by liuzeyang0311
  1

• The WER is always 1.000

  #13 opened 4 years ago by liuzeyang0311
  1

• Curriculum learning

  #12 opened 4 years ago by Byshev333
  3

• What does an iteration mean?

  #2 opened 4 years ago by kasri-mids
  3

• Broken pipe error in Dataloader

  #11 opened 4 years ago by smeetrs
  3

• tuple attribute error when running pretrain.py

  #9 opened 4 years ago by arunm95
  9

• Improvement in AV model performance over AO

  #5 opened 4 years ago by smeetrs
  13

• Details of the pretrained visual frontend and language models?

  #7 opened 4 years ago by kasri-mids
  1

• how much training time

  #4 opened 4 years ago by yuexianghubit
  12

• config.py file is missing?

  #1 opened 4 years ago by kasri-mids
  6