OOM on 4*L20（45G）when moving model from cpu to device NAVA/inference_nava.py", line 286, in main [rank0]: pipe = pipe.to(device)
，here‘s my running script：
CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun \ --nnodes=1 \ --nproc_per_node=4 \ --node_rank=0 \ --master_addr="127.0.0.1" \ --master_port=29507 \ inference_nava.py \ --config "./NAVA/nava.yaml" \ --ckpt "./NAVA/NAVA.ckpt" \ --out_dir "outputs" \ --data_format json \ --data_file "./NAVA/infer_cases/general/prompts.jsonl" \ --width 640 \ --height 480 \ --frames 37 \ --fps 24 \ --steps 50 \ --save_sample \ --gen_turn 1 \ --use_sp
how much VRAM is needed to run NAVA?
OOM on 4*L20（45G）when moving model from cpu to device
NAVA/inference_nava.py", line 286, in main [rank0]: pipe = pipe.to(device)，here‘s my running script：
CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun \ --nnodes=1 \ --nproc_per_node=4 \ --node_rank=0 \ --master_addr="127.0.0.1" \ --master_port=29507 \ inference_nava.py \ --config "./NAVA/nava.yaml" \ --ckpt "./NAVA/NAVA.ckpt" \ --out_dir "outputs" \ --data_format json \ --data_file "./NAVA/infer_cases/general/prompts.jsonl" \ --width 640 \ --height 480 \ --frames 37 \ --fps 24 \ --steps 50 \ --save_sample \ --gen_turn 1 \ --use_sphow much VRAM is needed to run NAVA?