Skip to content
This repository has been archived by the owner on Nov 3, 2023. It is now read-only.

[dist] Allow non-tcp based distributed setup. #3095

Merged
merged 5 commits into from
Sep 22, 2020
Merged

[dist] Allow non-tcp based distributed setup. #3095

merged 5 commits into from
Sep 22, 2020

Conversation

stephenroller
Copy link
Contributor

Patch description
See D23761809 for internal rationale of this patch.

In some settings, we may wish to use something other than tcp for distributed initialization. In particular, in internal cases, distributed initialization is supported by the env:// method. This patch refactors out the initialization method so that it is directly a parameter, rather than demanding the TCP based method.

Testing steps
See internal patch. CI here.

Copy link
Contributor

@EricMichaelSmith EricMichaelSmith left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Makes sense, good to have!

@stephenroller stephenroller merged commit 0ffd3a6 into master Sep 22, 2020
@stephenroller stephenroller deleted the patch1 branch September 22, 2020 19:26
mojtaba-komeili pushed a commit that referenced this pull request Sep 23, 2020
* master: (30 commits)
  Prevent MTurk config folders from needing __init__.py (#3105)
  [dist] Allow non-tcp based distributed setup. (#3095)
  Move torch.load/torch.save to PathManager. (#3094)
  Adding package dependabot couldn't resolve (#3104)
  Test AcceptabilityChecker (#3091)
  [DOC DAY] ISH OSS static turn annotations (#3053)
  Fix for flaky DialoGPT test (#3097)
  Temporary compatibility note (#3098)
  Update model_list.py (#3096)
  Adding features into the human+model turn annotation task (#3006)
  [Doc Day] Totally rewrite Worlds docs. (#3049)
  list trainable parameters (#3086)
  TaskMaster-2 (#2678)
  [tests] Switch to OverfitTeacher for some model tests. (#3055)
  Change drqa default tokenizer. (#3069)
  Self-chat batch size assertion error (#3081)
  Disable zipfile serialization option of torch.save (#3077)
  Always download data on interactive wizard. (#3079)
  Two inaugural FAQ questions (#3073)
  Add multiwoz v2.0 (#3072)
  ...
This is necessary in order to update the branch. It is outdated
and because of that a lot of its tests are failing.
stephenroller added a commit that referenced this pull request Sep 28, 2020
# v0.9.3 Release

Known issues
- Short options like `-m` and `-t` do fail in Python 3.8. Use `--model` and `--task`

Breaking Changes
- A number of old MTurk tasks have been archived and removed from the code (#3085)

New Features
- [image] Detectron feature extraction (#3083)
- [data] Natural questions (#3070)
- [data] TaskMaster-2 (#2678)
- [data] New versions of multiwoz (#3072)
- [distributed] Allow non-tcp based distributed setup (#3095)
- [core] Move torch.load/torch.save to PathManager. (#3094, #3077)
- [mturk] New task on static turn annotations (#3053)
- [mturk] New features in human+model annotation (#3006)
- [core] TorchClassifierAgent now prints its number of prameters (#3086)

Doc Changes:
- New Worlds tutorial (#3049)
- Tutorial on using `-t jsonfile` (#3061)
- Better help message for --init-model (#3090)
- Additions to FAQ (#3073)
- Updated model zoo descriptions for BlenderBot (#3096)

Bug Fixes
- Distributed evaluation now writes to world logs earlier (#3122)
- An argument was updated from store_true to bool (#3113)
- Self-chat now fails loudly with unexpected batchsize (#3081)
- Update drqa default tokenizer away from removed (#3069)
- Using wizard of wikipedia in interactive mode downloads data (#3079)

Developer notes:
- New pre-commit git-secrets (#3106)
- Code coverage improvements (#3110, #3091)
- More reliable tests. (#3108, #3097, #3055)
- Mephisto task dependencies have been updates due to security bugs (#3111, #3101, #3104)
- MTurk config folders are exempt from __init__.py requirements (#3105)
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants