owl-10
owl-10
> 设计的时候，没有考虑 stream_infer with session 的用法 但是如果想stream_infer能够多模态历史对话，采用openai-prompt也会出错
Hello, we will update the repo recently. And we will provide the code that train the vlm and decoder togather.
> > Hello, we will update the repo recently. And we will provide the code that train the vlm and decoder togather. > > Thank you for your answer. Does...
Same problem, recently appeared
> Hi! > > Quick question, my understanding is that ReCogDrive only uses the front view for driving, and the code says the same (seeing that the feature builder uses...
> `set -x > > TRAIN_TEST_SPLIT=mini > > export NUPLAN_MAP_VERSION="nuplan-maps-v1.0" export NUPLAN_MAPS_ROOT="/home/navsim/dataset/maps" export NAVSIM_EXP_ROOT="/home/navsim/exp" export NAVSIM_DEVKIT_ROOT="/home/recogdrive/navsim" export OPENSCENE_DATA_ROOT="/home/navsim/dataset/mini" > > export NCCL_IB_DISABLE=0 export NCCL_P2P_DISABLE=0 export NCCL_SHM_DISABLE=0 > > MASTER_PORT=${MASTER_PORT:-63669} PORT=${PORT:-63665}...
You should `pip install transformers==4.37.2`.
> Thanks, and how about huggingface-hub? I tried many versions but failed. Can you give me the error message you got?
> [@owl-10](https://github.com/owl-10) Excuse me again. How to train VLM on NVIDIA GeForce RTX 3090 with fp16? > > I encountered the following errors: > > ``` > [rank4]: Traceback (most...
> Hello Authors > > Can you please tell if this model is https://huggingface.co/owl10/ReCogDrive-VLM-DriveLM is specially trained on only DriveLM dataset? And what is the difference b/w this model and...