inference bug about open-sora

[lixirui142]

I solved the problem by adding the local host to "/etc/hosts" manually.

Try this: https://blog.csdn.net/lin_xiao_yi/article/details/132490694

[FrankLeeeee]

Hi, could you reproduce this issue with the following steps?

1. write a test.py

import torch

torch.distributed.init_process_group(backend="nccl")
print("test")                                                                                          

2. invoke this script with the following commands

torchrun --standalone --nproc_per_node 1 test.py

[lanson07]

Hi, could you reproduce this issue with the following steps?

1. write a test.py

import torch

torch.distributed.init_process_group(backend="nccl")
print("test")                                                                                          

2. invoke this script with the following commands

torchrun --standalone --nproc_per_node 1 test.py

i just try above, it runs normally.

[FrankLeeeee]

I see, that's weird. While I try to debug this, you can replace colossalai.launch() with the following lines first.

import torch
import torch.distributed as dist

dist.init_process_group(backend="nccl")
rank = dist.get_rank()
torch.cuda.set_device(rank)

[min-star]

I use the 4090 run the test.py,there still have the same error.

[maokangkun]

same problem, I have this warning

[2024-03-18 17:04:07,488] torch.distributed.run: [WARNING] master_addr is only used for static rdzv_backend and when rdzv_endpoint is not specified.

I try to add one line to modify the master_addr in inference.py:

os.environ["MASTER_ADDR"] = "127.0.0.1"
colossalai.launch_from_torch({})

but it didn't work

[Duyz232]

我也是遇到这种问题？不知如何解决

[Duyz232]

I see, that's weird. While I try to debug this, you can replace colossalai.launch() with the following lines first.

import torch
import torch.distributed as dist

dist.init_process_group(backend="nccl")
rank = dist.get_rank()
torch.cuda.set_device(rank)

我已经在anconda环境安装了
torch,但是不知道为什么还是找不到torch