[BUG] AttributeError: module 'deepspeed' has no attribute 'utils' when installing DeepSpeed from source

[KMFODA]

Describe the bug
When installing deepspeed from Github rather than using pip install deepspeed I get the below error. I've traced this back to commit 6ad3119d5420c6b25b456626a40f4ded8c05a81. Before this commit running these instructions works whereas from this commit onwards I get the below error:

Traceback (most recent call last):
  File "/usr/local/bin/deepspeed", line 3, in <module>
    from deepspeed.launcher.runner import main
  File "/usr/local/lib/python3.6/dist-packages/deepspeed/__init__.py", line 13, in <module>
    from . import ops
  File "/usr/local/lib/python3.6/dist-packages/deepspeed/ops/__init__.py", line 1, in <module>
    from . import adam
  File "/usr/local/lib/python3.6/dist-packages/deepspeed/ops/adam/__init__.py", line 1, in <module>
    from .cpu_adam import DeepSpeedCPUAdam
  File "/usr/local/lib/python3.6/dist-packages/deepspeed/ops/adam/cpu_adam.py", line 10, in <module>
    from deepspeed.utils.logging import should_log_le
  File "/usr/local/lib/python3.6/dist-packages/deepspeed/utils/__init__.py", line 1, in <module>
    from .logging import logger, log_dist
  File "/usr/local/lib/python3.6/dist-packages/deepspeed/utils/logging.py", line 5, in <module>
    import deepspeed.comm as dist
  File "/usr/local/lib/python3.6/dist-packages/deepspeed/comm/__init__.py", line 3, in <module>
    import deepspeed.utils as utils
AttributeError: module 'deepspeed' has no attribute 'utils'

To Reproduce
Steps to reproduce the behavior:

1. git clone https://github.com/microsoft/DeepSpeed/2.
2. cd DeepSpeed
3. pip install .
4. ds_report

Expected behavior
ds_report should run and print out a DeepSpeed environment report

ds_report output
N/A

Screenshots
N/A

System info (please complete the following information):

• OS = Linux 5.15.38
• GPU count = 4 TeslaV100S
• Python = Python 3.6.9

Launcher context
deepspeed launcher

Docker context
Are you using a specific docker image that you can share?

Additional context
Add any other context about the problem here.

[mrwyattii]

@KMFODA I've tried reproducing in a fresh environment, but I get no error. Could you try installing with a python virtual environment?

python -m virtualenv venv
source ./venv/bin/activate
git clone https://github.com/microsoft/DeepSpeed/ && cd DeepSpeed
pip install .
ds_report

[KMFODA]

I get the same error with a virtual environment unfortunately. If it helps I'm running this on a VM set up in ovhcloud. I think this might be an error related to how some of the subdirectories are being imported. I made these changes on my fork and they fixed this error.

Essential replacing imports such as:

import deepspeed.comm as dist
import deepspeed.runtime.lr_schedules as lr_schedules
import deepspeed.utils.groups as groups
import deepspeed.utils as utils

with:

from deepspeed import comm as dist
from deepspeed.runtime import lr_schedules
from deepspeed.utils import groups
from deepspeed import utils

not sure if these changes make sense?

[mrwyattii]

@KMFODA Would you be able to submit a PR from your modified fork?

[KMFODA]

yes of course just submitted #2026