torchrun command not found

[Micla-SHL]

Run:
bash dist_trigger_docker.sh hostfile Aquila-chat.yaml aquila-7b aquila_experiment
Error:

[INFO] bmtrain_mgpu.sh: hostfile configfile model_name exp_name exp_version
bmtrain_mgpu.sh: 行 84: torchrun：未找到命令

envs: 1 * 4090

torch                       2.1.0+cu118
torchaudio                  2.1.0
torchmetrics                1.2.0
torchvision                 0.16.0+cu118

hostfile: 192.168.1.5 slots=1
edit ~/FlagAI/flagai/env_args.py
self.parser.add_argument('--local_rank', default=0, type=int, help='start training from saved checkpoint')
be ineffective
What more can be done? I don't feel comfortable downgrading at this time. Are there any other options?

Originally posted by @Micla-SHL in #511 (comment)

[ftgreat]

use which torchrun to make sure whether torchrun is installed and can be found.

PS, could you tell us the purpose of using this script, pretraining or finetune? thanks

[Micla-SHL]

It depends on the script itself. My thought is to follow the readme step-by-step first, getting everything working, before considering any branching

[Micla-SHL]

Update:
I've just realized
dist_trigger_docker.sh # 多机多卡运行的脚本文件，单机可选用local_trigger_docker.sh
run:
bash local_trigger_docker.sh hostfile Aquila-chat.yaml aquila-7b aquila_experiment
Other errors were produced, so for now I'm assuming the torchrun error was skipped over.. torchrun command not found It's not a concern of mine anymore.

For users with multiple GPUs, manually setting slots=1 and running bash dist_trigger_docker.sh hostfile Aquila-chat.yaml aquila-7b aquila_experiment specifically will trigger a "torchrun command not found" error？

[Micla-SHL]

The problem that just came up:

  File "/Micla/Project/FlagAI/examples/Aquila/Aquila-chat/aquila_chat.py", line 15, in <module>
    from flagai.env_trainer_v1 import EnvTrainer
    ModuleNotFoundError: No module named 'flagai.env_trainer_v1'

~/examples/Aquila/Aquila-chat/aquila_chat.py

from flagai.auto_model.auto_loader import AutoLoader             #run ok
from flagai.data.tokenizer import Tokenizer                           #run ok
from flagai.env_args import EnvArgs
from flagai.env_trainer_v1 import EnvTrainer      ##run error No module named 'flagai.env_trainer_v1'
import jsonlines
import numpy as np
import cyg_conversation as conversation_lib
from flagai.model.tools.peft.prepare_lora import lora_transfer     ##No module named 'flagai.model.tools'

~/_init_.py is None, edit: from . import env_trainer_v1 be ineffective
edit: aquila_chat.py : import sys / sys.path.append("..") be ineffective

What more can be done?

[Micla-SHL]

Because https://github.com/FlagAI-Open/Aquila2 .issues close