运行 python 作业时，如何让 condor 自动导入我的 conda 环境？

Question

我正在向 condor 提交我的工作，但它说没有安装 tensorboard，这是错误的，因为我在一个交互式工作中运行，所以它已安装。

如何让 conda 使用我当前的活动 conda 环境？

我的神鹰提交脚本：

####################
#
# Experiments script
# Simple HTCondor submit description file
#
# reference: https://gitlab.engr.illinois.edu/Vision/vision-gpu-servers/-/wikis/HTCondor-user-guide#submit-jobs
#
# chmod a+x test_condor.py
# chmod a+x experiments_meta_model_optimization.py
# chmod a+x meta_learning_experiments_submission.py
# chmod a+x download_miniImagenet.py
#
# condor_submit -i
# condor_submit job.sub
#
####################

# Executable   = meta_learning_experiments_submission.py
# Executable = automl-proj/experiments/meta_learning/meta_learning_experiments_submission.py
# Executable = ~/automl-meta-learning/automl-proj/experiments/meta_learning/meta_learning_experiments_submission.py
Executable = /home/miranda9/automl-meta-learning/automl-proj/experiments/meta_learning/meta_learning_experiments_submission.py

## Output Files
Log          = condor_job.$(CLUSTER).log.out
Output       = condor_job.$(CLUSTER).stdout.out
Error        = condor_job.$(CLUSTER).err.out

# Use this to make sure 1 gpu is available. The key words are case insensitive.
REquest_gpus = 1
# requirements = ((CUDADeviceName = "Tesla K40m")) && (TARGET.Arch == "X86_64") && (TARGET.OpSys == "LINUX") && (TARGET.Disk >= RequestDisk) && (TARGET.Memory >= RequestMemory) && (TARGET.Cpus >= RequestCpus) && (TARGET.gpus >= Requestgpus) && ((TARGET.FileSystemDomain == MY.FileSystemDomain) || (TARGET.HasFileTransfer))
# requirements = (CUDADeviceName == "Tesla K40m")
# requirements = (CUDADeviceName == "Quadro RTX 6000")
requirements = (CUDADeviceName != "Tesla K40m")

# Note: to use multiple CPUs instead of the default (one CPU), use request_cpus as well
Request_cpus = 8

# E-mail option
Notify_user = me@gmail.com
Notification = always

Environment = MY_CONDOR_JOB_ID= $(CLUSTER)

# "Queue" means add the setup until this line to the queue (needs to be at the end of script).
Queue

我的提交脚本的前几行直到失败行：

#!/home/miranda9/.conda/bin/python3.7

import torch
import torch.nn as nn
import torch.optim as optim
# import torch.functional as F
from torch.utils.tensorboard import SummaryWriter

---

相关cmets：

我确实看到了这个问题how to run a python program on Condor? 和这个http://chtc.cs.wisc.edu/python-jobs.shtml，但我不敢相信我们必须这样做。集群中的其他所有人都不会做任何复杂的事情，而我之前已经运行了我的脚本，而不必做任何复杂的事情，我非常怀疑是否需要这样做。 p>

Answer 1

我真的不明白 condor 是如何工作的，但似乎一旦我将正确的 python 路径放在当前环境的顶部，它就开始工作了。所以检查你的python命令在哪里：

(automl-meta-learning) miranda9~/automl-meta-learning $ which python
~/miniconda3/envs/automl-meta-learning/bin/python

然后将其复制粘贴到 python 提交脚本的顶部：

#!/home/miranda9/miniconda3/envs/automl-meta-learning/bin/python

我希望我可以将所有这些都包含在 job.sub 中。如果你知道怎么做，请告诉我。

---

参考解决方案：https://stackoverflow.com/a/64484025/1601580

---

回声的克里斯蒂娜的解决方案

输入你的工作脚本：

getenv = True

我当前的提交脚本：

####################
#
# Experiments script
# Simple HTCondor submit description file
#
# reference: https://gitlab.engr.illinois.edu/Vision/vision-gpu-servers/-/wikis/HTCondor-user-guide#submit-jobs
#
# chmod a+x test_condor.py
# chmod a+x experiments_meta_model_optimization.py
# chmod a+x meta_learning_experiments_submission.py
# chmod a+x download_miniImagenet.py
# chmod a+x ~/meta-learning-lstm-pytorch/main.py
# chmod a+x /home/miranda9/automl-meta-learning/automl-proj/meta_learning/datasets/rand_fc_nn_vec_mu_ls_gen.py
# chmod a+x /home/miranda9/automl-meta-learning/automl-proj/experiments/meta_learning/supervised_experiments_submission.py
# chmod a+x /home/miranda9/automl-meta-learning/results_plots/is_rapid_learning_real.py
# chmod a+x /home/miranda9/automl-meta-learning/test_condor.py
# chmod a+x /home/miranda9/ML4Coq/main.sh
# chmod a+x /home/miranda9/ML4Coq/ml4coq-proj/PosEval/download_data.py
# chmod a+x /home/miranda9/ML4Coq/ml4coq-proj/pos_eval/create_pos_eval_dataset.sh
# chmod a+x /home/miranda9/ML4Coq/ml4coq-proj/embeddings_zoo/tree_nns/main_brando.py
# chmod a+x /home/miranda9/ML4Coq/main.sh
# condor_submit -i
# condor_submit job.sub
#
####################

# Executable = /home/miranda9/automl-meta-learning/automl-proj/experiments/meta_learning/supervised_experiments_submission.py

# Executable = /home/miranda9/automl-meta-learning/automl-proj/experiments/meta_learning/meta_learning_experiments_submission.py
# SUBMIT_FILE = meta_learning_experiments_submission.py

# Executable = /home/miranda9/meta-learning-lstm-pytorch/main.py
# Executable = /home/miranda9/automl-meta-learning/automl-proj/meta_learning/datasets/rand_fc_nn_vec_mu_ls_gen.py

# Executable = /home/miranda9/automl-meta-learning/results_plots/is_rapid_learning_real.py
# SUBMIT_FILE = is_rapid_learning_real.py

# Executable = /home/miranda9/automl-meta-learning/test_condor.py

# Executable = /home/miranda9/ML4Coq/ml4coq-proj/embeddings_zoo/tree_nns/main_brando.py
# SUBMIT_FILE = main_brando.py

# Executable = /home/miranda9/ML4Coq/ml4coq-proj/PosEval/download_data.py
# SUBMIT_FILE = ml4coq-proj/PosEval/download_data.py

# Executable = /home/miranda9/ML4Coq/ml4coq-proj/pos_eval/create_pos_eval_dataset.sh
# SUBMIT_FILE = create_pos_eval_dataset.sh

Executable = /home/miranda9/ML4Coq/main.sh
SUBMIT_FILE = main.sh

# Output Files
Log          = $(SUBMIT_FILE).log$(CLUSTER)
Output       = $(SUBMIT_FILE).o$(CLUSTER)
Error        = $(SUBMIT_FILE).o$(CLUSTER)

getenv = True
# cuda_version = 10.2
# cuda_version = 11.0

# Use this to make sure 1 gpu is available. The key words are case insensitive.
# REquest_gpus = 1
REquest_gpus = 2
requirements = (CUDADeviceName != "Tesla K40m")
requirements = (CUDADeviceName != "GeForce GTX TITAN X")
# requirements = (CUDADeviceName == "Quadro RTX 6000")
# requirements = ((CUDADeviceName != "Tesla K40m")) && (TARGET.Arch == "X86_64") && (TARGET.OpSys == "LINUX") && (TARGET.Disk >= RequestDisk) && (TARGET.Memory >= RequestMemory) && (TARGET.Cpus >= RequestCpus) && (TARGET.gpus >= Requestgpus) && ((TARGET.FileSystemDomain == MY.FileSystemDomain) || (TARGET.HasFileTransfer))
# requirements = (CUDADeviceName == "Tesla K40m")
# requirements = (CUDADeviceName == "GeForce GTX TITAN X")

# Note: to use multiple CPUs instead of the default (one CPU), use request_cpus as well
# Request_cpus = 1
Request_cpus = 4
# Request_cpus = 5
# Request_cpus = 8
# Request_cpus = 16
# Request_cpus = 32

# E-mail option
Notify_user = brando.science@gmail.com
Notification = always

Environment = MY_CONDOR_JOB_ID= $(CLUSTER)

# "Queue" means add the setup until this line to the queue (needs to be at the end of script).
Queue

Answer 2

HTCondor 在交互式和批处理作业中使用不同的默认环境。交互式作业复制与您的登录会话相同的 shell 环境（包括激活的 conda 环境）。批处理作业从一个非常精简的环境开始（要查看实际情况，请尝试使用/usr/bin/env 作为可执行文件运行测试作业）；激活的 conda 环境不会被带入批处理作业环境。

HTCondor 手册中描述了此行为和潜在的提交文件解决方案：https://htcondor.readthedocs.io/en/latest/users-manual/services-for-jobs.html?highlight=environment#environment-variables