[BUG] `pip install deepspeed==0.6.2` raises `ModuleNotFoundError: No module named 'op_builder'`

Describe the bug I’d like to report a possible bug in the installation of 0.6.2 in particular.

To Reproduce

$ # === from the host ===
$ docker pull pytorchlightning/pytorch_lightning:base-cuda-py3.7-torch1.11
$ docker run --rm -it --gpus all --ipc host pytorchlightning/pytorch_lightning:base-cuda-py3.7-torch1.11
$ # === within the container ===
$ pip list | grep -e deepspeed -e torch
deepspeed                     0.6.1
torch                         1.11.0+cu113
torchmetrics                  0.8.0
torchtext                     0.12.0
torchvision                   0.12.0+cu113
$ pip uninstall deepspeed -y
$ pip install deepspeed==0.6.2
Collecting deepspeed==0.6.2
  Downloading deepspeed-0.6.2.tar.gz (542 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 542.4/542.4 KB 19.5 MB/s eta 0:00:00
  Preparing metadata (setup.py) ... error
  error: subprocess-exited-with-error

  × python setup.py egg_info did not run successfully.
  │ exit code: 1
  ╰─> [6 lines of output]
      Traceback (most recent call last):
        File "<string>", line 36, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>
        File "/tmp/pip-install-brk5ad19/deepspeed_e004efb3362a460e85f320ebf0a47444/setup.py", line 35, in <module>
          from op_builder import ALL_OPS, get_default_compute_capabilities, OpBuilder
      ModuleNotFoundError: No module named 'op_builder'
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed

× Encountered error while generating package metadata.
╰─> See above for output.

note: This is an issue with the package mentioned above, not pip.
hint: See above for details.

FYI, installing other versions succeeds.

$ pip install deepspeed==0.6.1  # or other versions I've tried

Expected behavior Successful installation with no error raised.

pip install deepspeed==0.6.2

ds_report output Please run ds_report to give us details about your setup.

Screenshots Please refer to the commands above or to our CI run: https://dev.azure.com/PytorchLightning/pytorch-lightning/_build/results?buildId=69269&view=logs&j=3afc50db-e620-5b81-6016-870a6976ad29&t=96e63b50-ac1f-5745-9558-05f9481cd089&l=17

System info (please complete the following information):

• OS: Linux ad93fadeb85f 5.4.0-53-generic #59-Ubuntu SMP Wed Oct 21 09:38:44 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux
• GPU count and types [e.g. two machines with x8 A100s each]
• Interconnects (if applicable) [e.g., two machines connected with 100 Gbps IB]
• Python version: 3.7
• Any other relevant info about your setup

Launcher context n/a

Docker context

docker pull pytorchlightning/pytorch_lightning:base-cuda-py3.7-torch1.11

Additional context Encountered this issue while working on https://github.com/PyTorchLightning/pytorch-lightning/issues/12881.

[akihironitta]

Confirmed 0.6.3 works. Thank you for deploying the fix quickly.

[jeffra]

We’ve fixed the release and removed the 0.6.2 distribution since it does not install w/o error. Please install 0.6.3, please re-open if you’re still seeing the error but it should be resolved now.