Create your own GitHub profile
Sign up for your own profile on GitHub, the best place to host code, manage projects, and build software alongside 50 million developers.
Sign up
Popular repositories
-
Dependency-based Word Embeddings (Levy and Goldberg, 2014) with BZ2 compression support.
C 17
1,095 contributions in the last year
Activity overview
Contributed to
facebookresearch/ParlAI,
pytorch/pytorch,
pytorch/add-annotations-github-action
and 5 other
repositories
Contribution activity
May 2020
Created a pull request in facebookresearch/ParlAI that received 2 comments
Allow DialogTeacher's setup_data to emit non-tuple
Patch description Making DialogTeacher a bit more friendly for highly custom teachers, which have more than just text/label. Built in some safety c…
+139
−75
•
2
comments
- Testing for Unigram and Starspace agents
- Automatically mark stale issues/PRs
- Fix broken link in chat service tutorial.
- Ignore F541 f-string is missing placeholders
- Teacher-specific evaluations
- TaskMaster-2
- Add Opt load/save utilities
- Include blender image with `python setup.py install`
- Update README.md
- Force PyTorch 1.4.0
- Compliance with internal OSS policies.
- Enable pytorch 1.5.
- Party parrot
- Fix off by one in build_data.
- Update NEWS.md
- Support add_argument(type=bool)
- Improved logging and show alternative beams.
- ELI5, Download pushshift Reddit and CC Dumps
- [Scaling] Background preprocessing
- Global parlai command line launcher
- Fixed Sphinx :return: documentation
- Use packaging.version to parse PyTorch version.
- Add classification metrics
- Added --host and --port parameter to client.py and interactive_web.py
- [DodecaDialogue] Dodeca oss
- [Dodeca] Image Chat Generation Task
- Add some more links in README
- [Interactive] Fix for single turn
- Label to text teacher
- Use safe personas in interactive BST
- Restore steamrollered colors, via style choice
- Remove extra "if you" from README.md
- Update PR template
- Updates to dialogue unlikelihood project
Created an issue in pytorch/pytorch that received 1 comment
clip_grad_norm_ fails when parameters are on different devices (1.5 regression)
clip_grad_norm_ worked even when parameters were on different devices. Pytorch 1.5 no longer supports this, due to #32020.
To…
1
comment
11
contributions
in private repositories
May 7 – May 21