Contributing Guide (#1138)

* contributing guide * Update NeoXArgs docs automatically * Update CONTRIBUTING.md * Update NeoXArgs docs automatically * Remove microsoft references and link on main readme * Update NeoXArgs docs automatically * pre-commit * Update NeoXArgs docs automatically --------- Co-authored-by: github-actions <[email protected]> Co-authored-by: Quentin Anthony <[email protected]>
EleutherAI · Jan 29, 2024 · e5602c3 · e5602c3
1 parent 3d8fec0
commit e5602c3
Show file tree

Hide file tree

Showing 5 changed files with 104 additions and 4 deletions.
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -16,10 +16,10 @@ repos:
  args: [--fix=lf]
  - id: requirements-txt-fixer
  - id: trailing-whitespace
- - repo: https://gitlab.com/daverona/pre-commit-cpp
+ - repo: https://gitlab.com/daverona/pre-commit/cpp
  rev: 0.8.0
  hooks:
- - id: clang-format # formatter of C/C++ code based on a style guide: LLVM, Google, Chromium, Mozilla, and WebKit available
+ - id: clang-format  # formatter of C/C++ code based on a style guide: LLVM, Google, Chromium, Mozilla, and WebKit available
  args: []
 
  - repo: https://github.com/psf/black

diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -0,0 +1,86 @@
+# Contributing
+GPT-NeoX welcomes your contributions!
+
+## Prerequisites
+GPT-NeoX uses [pre-commit](https://pre-commit.com/) to ensure that formatting is
+consistent across GPT-NeoX. First, ensure that `pre-commit` is installed with
+`pip install pre-commit`. Next, the pre-commit hooks must be installed once
+before commits can be made:
+```bash
+pre-commit install
+```
+Please install `clang-format` from Conda:
+```bash
+conda install clang-format
+```
+
+Afterwards, our suite of formatting tests run automatically before each `git commit`. You
+can also run these manually:
+```bash
+pre-commit run --all-files
+```
+If a formatting test fails, it will fix the modified code in place and abort
+the `git commit`. After looking over the changes, you can `git add <modified files>`
+and then repeat the previous `git commit` command.
+
+
+## Testing
+GPT-NeoX tracks two types of tests: unit tests and more costly model convergence tests.
+Unit tests are found in `tests/unit/` and the model convergence tests are found in
+`tests/model/`.
+
+### Unit Tests
+[PyTest](https://docs.pytest.org/en/latest/) is used to execute tests. PyTest can be
+installed from PyPI via `pip install pytest`. Simply invoke `pytest --forked` to run the
+unit tests:
+```bash
+pytest --forked tests/unit/
+```
+You can also provide the `-v` flag to `pytest` to see additional information about the
+tests. Note that [pytest-forked](https://github.com/pytest-dev/pytest-forked) and the
+`--forked` flag are required to test CUDA functionality in distributed tests.
+
+### Model Tests
+To execute model tests, first install GPT-NeoX. Next, execute the model test driver:
+```bash
+cd tests/model/
+pytest run_sanity_check.py
+```
+Note that the `--forked` flag is not necessary for the model tests.
+
+## Contributor License Agreement
+This project welcomes contributions and suggestions. Most contributions require you to
+agree to a Contributor License Agreement (CLA) declaring that you have the right to, and
+actually do, grant us the rights to use your contribution. For details, visit
+https://cla-assistant.io/EleutherAI/gpt-neox.
+
+When you submit a pull request, a CLA bot will automatically determine whether you need
+to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply
+follow the instructions provided by the bot. You will only need to do this once across
+all repos using our CLA.
+
+## New Feature Contribution Guidelines
+Unlike bug fix or improving existing feature (where users usually directly submit a PR and we review it), adding a new feature to GPT-NeoX requires several steps: (1) proposal and discussion, (2) implementation and verification, (3) release and maintenance. This general guideline applies to all new feature contributions. Core GPT-NeoX team member contributions may complete step 1 internally.
+
+### Step 1: Proposal and Discussion
+We ask users to first post your intended feature in an issue. This issue needs to include:
+
+* A description of the proposed feature.
+* A motivation of why it will be useful to GPT-NeoX users.
+* A rough design of how you implement the feature inside GPT-NeoX.
+* (Important) Results or planned experiments to demonstrate the effectiveness and correctness of the feature.
+ * If the feature only affects performance and does not affect training convergence, we require testing on a fraction of training to demonstrate that the training/validation loss are consistent with baseline, and that the performance is better than baseline.
+ * If the feature does affect training convergence, we require testing the whole training to demonstrate that the feature achieves better/on-par final model quality and training performance compared to baseline.
+
+Based on the issue we shall discuss the merit of the new feature and decide whether to accept or decline the proposal. Once accepted and after we confirm the design and implementation plan, we are ready for step 2.
+
+### Step 2: Implementation and Verification
+The contributor will proceed and implement the feature, and the GPT-NeoX team will provide guidance/helps as needed. The required deliverables include:
+
+* A PR to [EleutherAI/GPT-NeoX](https://github.com/EleutherAI/gpt-neox) including (1) the feature implementation (2) unit tests (3) documentation (4) example usage.
+* In the implementation (code, documentation, tutorial), we require the feature author to record their GitHub username as a contact method for future questions/maintenance.
+
+After receiving the PRs, we will review them and merge them after necessary tests/fixes.
+
+### Step 3: Release and Maintenance
+After the PRs are merged, we will announce the feature on our website (with credit to the feature author). We ask the feature author to commit to the maintenance of the feature.
diff --git a/README.md b/README.md
@@ -76,6 +76,7 @@ Prior to 3/9/2023, GPT-NeoX relied on [DeeperSpeed](https://github.com/EleutherA
  + [Other Modalities](#other-modalities)
 - [Administrative Notes](#administrative-notes)
  * [Citing GPT-NeoX](#citing-gpt-neox)
+ * [Contributing](#contributing)
  * [Licensing](#licensing)
  * [Acknowledgements](#acknowledgements)
 
@@ -637,6 +638,11 @@ To cite the 20 billion parameter model named `GPT-NeoX-20B`, please use
 }
 ```
 
+## Contributing
+GPT-NeoX is built by the open-source AI community, and relies on our amazing contributors! Please see our
+[contributing](CONTRIBUTING.md) guide for more details on our CLA, code formatting, testing,
+etc.
+
 ## Licensing
 
 This repository hosts code that is part of EleutherAI's GPT-NeoX project. Copyright (c) 2024, EleutherAI. Licensed under the Apache License:

diff --git a/configs/neox_arguments.md b/configs/neox_arguments.md
@@ -111,7 +111,7 @@ Logging Arguments
 
 - **git_hash**: str
 
- Default = 765c219
+ Default = cd9a8ea
 
  current git hash of repository
 
@@ -261,6 +261,14 @@ Model Arguments
 
 
 
+- **layernorm_fusion**: bool
+
+ Default = False
+
+ Use fused layer norm kernel (if `norm` is `layernorm`).
+
+
+
 - **use_qk_layernorm**: bool
 
  Default = False

diff --git a/megatron/neox_arguments/arguments.py b/megatron/neox_arguments/arguments.py
@@ -510,7 +510,7 @@ def get_deepspeed_main_args(self):
  args_list.extend(
  self.convert_key_value_to_command_line_arg("account", account)
  )
- 
+
  # master_address = os.environ['SLURM_JOB_NODELIST'].split('\n')[0]
  # args_list.extend(
  # self.convert_key_value_to_command_line_arg('master_addr', master_address)