search

ViTAE-Transformer / DeepSolo

The official repo for [CVPR'23] "DeepSolo: Let Transformer Decoder with Explicit Points Solo for Text Spotting" & [ArXiv'23] "DeepSolo++: Let Transformer Decoder with Explicit Points Solo for Multilingual Text Spotting"
Other
242 stars 34 forks source link

num_gpu #60

Open jihuan1203 opened 7 months ago

jihuan1203 commented 7 months ago

我添加了：

    os.environ["CUDA_VISIBLE_DEVICES"] = "1,2"

其中，指定

            "args": [
                "--config-file",
                "configs/FPN_R_50/CTW1500/finetune_96voc_50maxlen.yaml",
                "--num-gpus",
                "2"
            ],

在train_net.py中，并且torch.cuda.device_count() = 2 报错：

  File "/home/wit/anaconda3/envs/lj_py3.8pt1.10/lib/python3.8/site-packages/detectron2/engine/launch.py", line 119, in _distributed_worker
    assert num_gpus_per_machine <= torch.cuda.device_count()
AssertionError

想知道num_gpus_per_machine 值与什么因素有关