mlvlab/Flipped-VQA

Issues

• How was the STAR dataset preprocessed for this code

  #19 opened 2 months ago by SimMLPhys
  1

• What are the average stats being reported at the end of every epoch in training?

  #18 opened 4 months ago by SimMLPhys
  2

• Concerns and Clarifications Regarding MCQ to Generation Task Conversion

  #10 opened 4 months ago by inesriahi
  3

• How many GPUs are needed to train the model?

  #9 opened 4 months ago by jianhua2022
  2

• A question about QAV task in the code

  #8 opened 4 months ago by inesriahi
  2

• What is the function of the parameter `max_feats`?

  #7 opened 4 months ago by kuhne12
  1

• Checkpoints

  #5 opened 4 months ago by iquibalh
  1

• From where to download LLaMA-v1 model?

  #4 opened 4 months ago by inesriahi
  3

• How to use a trained checkpoint to make inference on validation set and resume from checkpoint.

  #3 opened 4 months ago by iquibalh
  5

• Error when training with TVQA dataset: AttributeError in DataLoader worker process

  #2 opened 4 months ago by iquibalh
  1

• How to extract features using CLIP VIT-L?

  #1 opened 4 months ago by yuanrr
  2

• Number of frames and its use in code and max_feats10 for video feature

  #14 opened 4 months ago by kumarmanas
  1

• need llama-13B finetuned checkpoints

  #17 opened 5 months ago by Huangbukun
  0

• Not getting the reported number.

  #11 opened 6 months ago by a6o
  4

• about self.gate2

  #16 opened 6 months ago by dunknsabsw
  1

• meaning of qav loss

  #15 opened 6 months ago by dunknsabsw
  1

• finetuned using lamma-13B

  #13 opened 6 months ago by Huangbukun
  3

• Cannot reproduce the result

  #12 opened 6 months ago by pseudo-aloha
  2

• ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 3 (pid: 55662) of binary: /usr/bin/python3

  #6 opened 7 months ago by inesriahi
  3