magic-research/PLLaVA

Issues

• TypeError: PllavaProcessor.__init__() got an unexpected keyword argument 'image_token'

  #87 opened a month ago
  1

• The download link has expired

  #86 opened 2 months ago
  1

• how to train train_tcllava_7b.sh on multi machines?

  #85 opened 3 months ago
  0

• flash-attn

  #84 opened 3 months ago
  0

• Does PLLAVA possess the capability to follow instructions?

  #83 opened 4 months ago
  0

• TGIF-QA evaluation crashing

  #82 opened 4 months ago
  3

• why the output is blank

  #81 opened 4 months ago
  4

• Why is the answer to this demo blank

  #80 opened 4 months ago
  2

• environment failed

  #79 opened 5 months ago
  0

• Image token process malfunction

  #78 opened 5 months ago
  0

• No module named dataset

  #76 opened 6 months ago
  1

• Is the inference done using only one <image> token?

  #75 opened 6 months ago
  3

• Question about the training data

  #74 opened 6 months ago
  0

• Eval Results.

  #73 opened 6 months ago
  6

• Training dataset keys_indexfile

  #72 opened 6 months ago
  0

• Training Dataset

  #71 opened 6 months ago
  1

• High video loss

  #70 opened 6 months ago
  1

• Which train_corpus was used by train pllava-13b?

  #69 opened 6 months ago
  0

• Instruction follow ability is weak

  #68 opened 7 months ago
  7

• [TRAINING DATA]About the number of video_qa in magic_jsons, which is more than the reported number in PLLaVA paper!

  #67 opened 6 months ago
  0

• Training loss curve

  #66 opened 2 months ago
  2

• The load_video function does not seem right in eval_utils.py

  #65 opened 7 months ago
  0

• What is the prompt used to evaluate models answer on VideoChatGPT generation benchmark?

  #64 opened 7 months ago
  0

• The inference result is all like a mess

  #63 opened 7 months ago
  3

• Inference error

  #61 opened 7 months ago
  4

• Hi, how does the <image> being replaced in PLLava

  #60 opened 7 months ago
  1

• Abou the dataset

  #59 opened 7 months ago
  0

• About the lora config alpha and rank

  #58 opened 7 months ago
  2

• How to prepare eval dataset.

  #57 opened 7 months ago
  4

• How to fintune base on your fintuned result

  #56 opened 7 months ago
  2

• The Avg. results in the PLLaVA paper is wrong!

  #55 opened 7 months ago
  2

• How to train only the projector?

  #54 opened 7 months ago
  2

• What's the 7B model score on mvbench?

  #53 opened 7 months ago
  1

• Difference of videochat2 jsons and magic jsons.

  #52 opened 7 months ago
  2

• Training time consumption

  #51 opened 7 months ago
  2

• eval on fintuning model

  #50 opened 7 months ago
  8

• Checkpoints for trained models?

  #49 opened 7 months ago
  1

• Multi-GPU Inference for 34B Model

  #48 opened 7 months ago
  1

• Clarifying Ambiguity in Training Data

  #47 opened 7 months ago
  3

• How can we quantize the 34B model for inference?

  #46 opened 8 months ago
  3

• No output text when evaluating on my own pre-trained model.

  #45 opened 7 months ago
  5

• It seems that the most effective point of increase is to increase the size of llm

  #44 opened 8 months ago
  0

• finetune problem

  #43 opened 8 months ago
  3

• Possible error in doctring for "pixel_values"

  #42 opened 8 months ago
  0

• Bug in tasks/eval/eval_utils.py

  #40 opened 8 months ago
  1

• Finetune my own's tasks based on pllava-7b, but got "image_to_overwrite.sum() != image_features.shape[:-1].numel()" assertion failure.

  #39 opened 8 months ago
  1

• Can PLLaVA caption rectangular videos without cropping?

  #38 opened 8 months ago
  1

• CLI inference

  #37 opened 8 months ago
  0

• Dataset Resquest

  #36 opened 8 months ago
  2

• About vram and int-4 version

  #35 opened 8 months ago
  0