-
To use EgoVLPv2 (specifically EgoTaskQA) on custom VQA dataset,
it is essential to preprocess the metadata about questions ans answers.
Specifically the "answer_encode" is needed.
How can I encod…
-
Hi,
I am trying to run the code from the KGen_VQA repository, but I encountered an issue due to the missing 'aokvqa_val_kb.pkl' and 'a-okvqa-repre.pkl' files. The code requires these files to execu…
-
I was reading your code and noticed something strange in the VSMForCausalLM class, maybe a small bug, but not sure since I haven't tested your code yet.
Why does images_clip (preprocessed by CLIPProc…
-
I have some doubts about the vqa_score, could you provide some more detailed information?
-
hi, thank you for your open source. But this code can't use for blip2. Do you have the code for the blip2 version?
-
Appreciate your brilliant work!
I'm much curious about your processing pipeline for splitting frames from Cholec80 videos, could you share the scripts with us?
Thanks
-
hello.
i had tried the vision fine-tuning script for `glm-4v-9b` model. The command i had used was `python3 finetune_demo/finetune_vision.py ./data THUDM/glm-4v-9b ./finetune_demo/configs/lora.yaml…
-
step1
pretrain_projector_image_encoder.sh
step2
pretrain_projector_video_encoder.sh
step3
finetune_dual_encoder.sh
step4
eval/vcgbench/inference/run_ddp_inference.sh
step5
eval/vcgbench/gpt_e…
-
Hi, I'm trying to use your Infer for a single MP4 video cmd. However, no matter which model I chose, e.g., FasterVQA, Fast-VQA-M, it raises error: `Segmentation fault (core dumped)`. I tried to debug …
-
In the vqa_dataset.py, I don't understand when the split != 'train', self.annotation will be set to 'color_test.json'. Should we modify this name manually when we are testing 'shape' or 'texture'?
!…
k101w updated
6 months ago