[chores] Rename master to main everywhere (#190)

Summary:
Pull Request resolved: fairinternal/mmf-internal#190

Pull Request resolved: #1121

Renaming master references to main at most places as part of terminology shift.

Reviewed By: ebsmothers

Differential Revision: D31623714

fbshipit-source-id: a8052d9606373f5bfb7c9ad6bb78628bbfa220e7

Author: apsdehal

.github/CONTRIBUTING.md

@@ -40,7 +40,7 @@ We take the following factors into consideration before accepting features and P
 ### Process
 
 1. Read the [PR guidelines](#guidelines) if you haven't.
-1. Fork the repo and create your branch from `master`.
+1. Fork the repo and create your branch from `main`.
 1. If your PR contains multiple orthogonal changes, split it to several PRs. Keep one PR focused on a single change while keeping it small.
 1. If you've added code that should be tested, add tests.
 1. Follow the [coding style guidelines](#coding-style) mentioned below.
@@ -64,7 +64,7 @@ pip install pre-commit && pre-commit install
 
 After this pre-commit hooks will be run before every commit.
 
-* Read the [editorconfig](https://github.com/facebookresearch/mmf/blob/master/.editorconfig) file to understand the exact coding style preferences.
+* Read the [editorconfig](https://github.com/facebookresearch/mmf/blob/main/.editorconfig) file to understand the exact coding style preferences.
 
 * Ideally, black and isort should be run via pre-commit hooks.
 But if for some reason you want to run black and isort separately follow this:

.github/PULL_REQUEST_TEMPLATE.md

@@ -3,4 +3,4 @@ Thanks for your contribution!
 If you're sending a large PR (e.g., >50 lines), please open an issue first about
 the feature/bug, and indicate how you want to contribute.
 
-Use [contributing guidelines](https://github.com/facebookresearch/mmf/tree/master/.github/CONTRIBUTING.md) before opening up the PR to follow MMF style guidelines.
+Use [contributing guidelines](https://github.com/facebookresearch/mmf/tree/main/.github/CONTRIBUTING.md) before opening up the PR to follow MMF style guidelines.

.github/workflows/deploy_website.yaml

@@ -3,21 +3,21 @@ name: Deploy Website 🚀
 on:
   push:
     branches:
-      - master
+      - main
 jobs:
   build:
     runs-on: ubuntu-latest
     steps:
-    - name: Checkout master 🛎️
+    - name: Checkout main 🛎️
       uses: actions/checkout@v2
       # If you're using actions/checkout@v2 you must set persist-credentials to false in most
       # cases for the deployment to work correctly.
       with:
         persist-credentials: false
         # 0 here indicates fetch all history for proper doc setup
         fetch-depth: 0
-        ref: master
-        path: mmf_master
+        ref: main
+        path: mmf_main
 
     - name: Setup Miniconda
       uses: conda-incubator/setup-miniconda@v2
@@ -53,7 +53,7 @@ jobs:
       shell: bash -l {0}
       run: |
         conda activate mmf
-        cd ${GITHUB_WORKSPACE}/mmf_master
+        cd ${GITHUB_WORKSPACE}/mmf_main
         python setup.py install
         python -c 'import torch; print("Torch version:", torch.__version__)'
         python -m torch.utils.collect_env
@@ -74,4 +74,4 @@ jobs:
       uses: peaceiris/actions-gh-pages@v3
       with:
         github_token: ${{ secrets.GITHUB_TOKEN }}
-        publish_dir: ./mmf_master/website/build
+        publish_dir: ./mmf_main/website/build

.pre-commit-config.yaml

@@ -10,6 +10,8 @@ repos:
     -   id: trailing-whitespace
     -   id: check-ast
     -   id: check-merge-conflict
+    -   id: no-commit-to-branch
+        args: ['--branch=main']
     -   id: no-commit-to-branch
         args: ['--branch=master']
     -   id: check-added-large-files

README.md

@@ -18,7 +18,7 @@
 
 MMF is a modular framework for vision and language multimodal research from Facebook AI Research. MMF contains reference implementations of state-of-the-art vision and language models and has powered multiple research projects at Facebook AI Research. See full list of project inside or built on MMF [here](https://mmf.sh/docs/notes/projects).
 
-MMF is powered by PyTorch, allows distributed training and is un-opinionated, scalable and fast. Use MMF to **_bootstrap_** for your next vision and language multimodal research project by following the [installation instructions](https://mmf.sh/docs/). Take a look at list of MMF features [here](https://mmf.sh/docs/getting_started/features).
+MMF is powered by PyTorch, allows distributed training and is un-opinionated, scalable and fast. Use MMF to **_bootstrap_** for your next vision and language multimodal research project by following the [installation instructions](https://mmf.sh/docs/getting_started/installation). Take a look at list of MMF features [here](https://mmf.sh/docs/getting_started/features).
 
 MMF also acts as **starter codebase** for challenges around vision and
 language datasets (The Hateful Memes, TextVQA, TextCaps and VQA challenges). MMF was formerly known as Pythia. The next video shows an overview of how datasets and models work inside MMF. Checkout MMF's [video overview](https://mmf.sh/docs/getting_started/video_overview).

docs/source/_templates/theme_variables.jinja

@@ -1,8 +1,8 @@
 {%-
 set external_urls = {
   'github': 'https://github.com/facebookresearch/mmf',
-  'github_issues': 'https://github.com/facebookresearch/master/issues',
-  'contributing': 'https://github.com/facebookresearch/master/blob/master/CONTRIBUTING.md',
+  'github_issues': 'https://github.com/facebookresearch/mmf/issues',
+  'contributing': 'https://github.com/facebookresearch/mmf/blob/main/CONTRIBUTING.md',
   'api': 'https://mmf.sh/api',
   'docs': 'https://mmf.sh/docs',
   'previous_pytorch_versions': 'https://mmf.sh/previous-versions/',

docs/source/conf.py

@@ -65,8 +65,8 @@
 source_suffix = [".rst", ".md"]
 # source_suffix = '.rst'
 
-# The master toctree document.
-master_doc = "index"
+# The main toctree document.
+main_doc = "index"
 
 # General information about the project.
 project = "MMF"
@@ -167,15 +167,15 @@
 # (source start file, target name, title,
 #  author, documentclass [howto, manual, or own class]).
 latex_documents = [
-    (master_doc, "mmf.tex", "MMF Documentation", "Facebook AI Research", "manual")
+    (main_doc, "mmf.tex", "MMF Documentation", "Facebook AI Research", "manual")
 ]
 
 
 # -- Options for manual page output ---------------------------------------
 
 # One entry per manual page. List of tuples
 # (source start file, name, description, authors, manual section).
-man_pages = [(master_doc, "mmf", "MMF Documentation", [author], 1)]
+man_pages = [(main_doc, "mmf", "MMF Documentation", [author], 1)]
 
 
 # -- Options for Texinfo output -------------------------------------------
@@ -185,7 +185,7 @@
 #  dir menu entry, description, category)
 texinfo_documents = [
     (
-        master_doc,
+        main_doc,
         "mmf",
         "MMF Documentation",
         author,
@@ -195,7 +195,7 @@
     )
 ]
 
-github_doc_root = "https://github.com/facebookresearch/mmf/tree/master"
+github_doc_root = "https://github.com/facebookresearch/mmf/tree/main"
 
 
 # At the bottom of conf.py

docs/source/index.rst

@@ -1,4 +1,4 @@
-.. mmf documentation master file, created by
+.. mmf documentation main file, created by
    sphinx-quickstart on Tue Apr 23 10:42:55 2019.
    You can adapt this file completely to your liking, but it should at least
    contain the root `toctree` directive.

mmf/common/test_reporter.py

@@ -11,7 +11,7 @@
 from mmf.common.registry import registry
 from mmf.common.sample import convert_batch_to_sample_list
 from mmf.utils.configuration import get_mmf_env
-from mmf.utils.distributed import gather_tensor, is_master
+from mmf.utils.distributed import gather_tensor, is_main
 from mmf.utils.file_io import PathManager
 from mmf.utils.general import ckpt_name_from_core_args, foldername_from_config_override
 from mmf.utils.logger import log_class_usage
@@ -115,7 +115,7 @@ def next_dataset(self, flush_report=True):
             return True
 
     def flush_report(self):
-        if not is_master():
+        if not is_main():
             # Empty report in all processes to avoid any leaks
             self.report = []
             return

mmf/datasets/base_dataset_builder.py

@@ -32,14 +32,13 @@ def load(self, config, dataset_type, *args, **kwargs):
         def build(self, config, dataset_type, *args, **kwargs):
             ...
 
-.. _here: https://github.com/facebookresearch/mmf/blob/master/mmf/datasets/vqa/vqa2/builder.py
+.. _here: https://github.com/facebookresearch/mmf/blob/main/mmf/datasets/vqa/vqa2/builder.py
 """
 import uuid
 from typing import Optional
 
 import pytorch_lightning as pl
 from mmf.utils.build import build_dataloader_and_sampler
-from mmf.utils.distributed import is_master, synchronize
 from mmf.utils.logger import log_class_usage
 from omegaconf import DictConfig
 from torch.utils.data import Dataset
@@ -77,14 +76,14 @@ def dataset_name(self, dataset_name):
 
     def prepare_data(self, config, *args, **kwargs):
         """
-        NOTE: The caller to this function should only call this on master process
+        NOTE: The caller to this function should only call this on main process
         in a distributed settings so that downloads and build only happen
-        on master process and others can just load it. Make sure to call
+        on main process and others can just load it. Make sure to call
         synchronize afterwards to bring all processes in sync.
 
         Lightning automatically wraps datamodule in a way that it is only
-        called on a master node, but for extra precaution as lightning
-        can introduce bugs, we should always call this under master process
+        called on a main node, but for extra precaution as lightning
+        can introduce bugs, we should always call this under main process
         with extra checks on our sides as well.
         """
         self.config = config
@@ -129,9 +128,9 @@ def build_dataset(self, config, dataset_type="train", *args, **kwargs):
         time when it is not available. This internally calls 'build' function.
         Override that function in your child class.
 
-        NOTE: The caller to this function should only call this on master process
+        NOTE: The caller to this function should only call this on main process
         in a distributed settings so that downloads and build only happen
-        on master process and others can just load it. Make sure to call
+        on main process and others can just load it. Make sure to call
         synchronize afterwards to bring all processes in sync.
 
         Args:

mmf/datasets/builders/clevr/dataset.py

@@ -3,10 +3,9 @@
 
 import numpy as np
 import torch
-from mmf.common.registry import registry
 from mmf.common.sample import Sample
 from mmf.datasets.base_dataset import BaseDataset
-from mmf.utils.distributed import is_master, synchronize
+from mmf.utils.distributed import is_main, synchronize
 from mmf.utils.general import get_mmf_root
 from mmf.utils.text import VocabFromText, tokenize
 from PIL import Image
@@ -81,7 +80,7 @@ def load(self):
             self.questions = json.load(f)[_CONSTANTS["questions_key"]]
 
             # Vocab should only be built in main process, as it will repetition of same task
-            if is_master():
+            if is_main():
                 self._build_vocab(self.questions, _CONSTANTS["question_key"])
                 self._build_vocab(self.questions, _CONSTANTS["answer_key"])
             synchronize()

mmf/datasets/builders/okvqa/dataset.py

@@ -2,13 +2,11 @@
 from typing import Type, Union
 
 import torch
-import tqdm
 from mmf.common.sample import Sample
 from mmf.common.typings import MMFDatasetConfigType
 from mmf.datasets.builders.okvqa.database import OKVQAAnnotationDatabase
 from mmf.datasets.mmf_dataset import MMFDataset
 from mmf.datasets.processors import GraphVQAAnswerProcessor
-from mmf.utils.distributed import is_master
 
 
 class OKVQADataset(MMFDataset):

mmf/datasets/builders/vqa2/dataset.py

@@ -5,7 +5,7 @@
 import tqdm
 from mmf.common.sample import Sample
 from mmf.datasets.mmf_dataset import MMFDataset
-from mmf.utils.distributed import is_master
+from mmf.utils.distributed import is_main
 
 
 logger = logging.getLogger(__name__)
@@ -42,7 +42,7 @@ def try_fast_read(self):
             )
             self.cache = {}
             for idx in tqdm.tqdm(
-                range(len(self.annotation_db)), miniters=100, disable=not is_master()
+                range(len(self.annotation_db)), miniters=100, disable=not is_main()
             ):
                 self.cache[idx] = self.load_item(idx)
 

mmf/datasets/databases/features_database.py

@@ -5,7 +5,7 @@
 import tqdm
 from mmf.datasets.databases.image_database import ImageDatabase
 from mmf.datasets.databases.readers.feature_readers import FeatureReader
-from mmf.utils.distributed import is_master
+from mmf.utils.distributed import is_main
 from mmf.utils.general import get_absolute_path
 
 
@@ -47,7 +47,7 @@ def _threaded_read(self):
         elements = [idx for idx in range(1, len(self.annotation_db))]
         pool = ThreadPool(processes=4)
 
-        with tqdm.tqdm(total=len(elements), disable=not is_master()) as pbar:
+        with tqdm.tqdm(total=len(elements), disable=not is_main()) as pbar:
             for i, _ in enumerate(pool.imap_unordered(self._fill_cache, elements)):
                 if i % 100 == 0:
                     pbar.update(100)

mmf/datasets/multi_dataset_loader.py

@@ -7,7 +7,6 @@
 import warnings
 from typing import Dict, Iterator
 
-import numpy as np
 import torch
 from mmf.common.sample import SampleList, convert_batch_to_sample_list
 from mmf.datasets import iteration_strategies
@@ -17,7 +16,7 @@
     broadcast_scalar,
     get_world_size,
     is_dist_initialized,
-    is_master,
+    is_main,
     is_xla,
 )
 from mmf.utils.general import get_batch_size, get_current_device
@@ -47,7 +46,7 @@ def __init__(
 
         self._iteration_strategy = iteration_strategy
         self._loaders = loaders
-        self._is_master = is_master()
+        self._is_main = is_main()
         self._num_datasets = len(self.loaders)
         self.dataset_list = list(loaders.keys())
         self._iterators = {}
@@ -230,7 +229,7 @@ def change_dataloader(self):
             self.current_index = choice
             return
 
-        if self._is_master:
+        if self._is_main:
             choice = self.iteration_strategy()
 
             # self._finished_iterators will always be empty in case of

mmf/datasets/processors/processors.py

@@ -82,7 +82,7 @@ def __call__(self, item, *args, **kwargs):
 from mmf.common.registry import registry
 from mmf.common.typings import ProcessorConfigType
 from mmf.utils.configuration import get_mmf_cache_dir, get_mmf_env
-from mmf.utils.distributed import is_master, synchronize
+from mmf.utils.distributed import is_main, synchronize
 from mmf.utils.file_io import PathManager
 from mmf.utils.logger import log_class_usage
 from mmf.utils.text import VocabDict
@@ -421,15 +421,15 @@ def __init__(self, config, *args, **kwargs):
             self._try_download()
 
     def _try_download(self):
-        _is_master = is_master()
+        _is_main = is_main()
 
         if self._already_downloaded:
             return
 
         needs_download = False
 
         if not hasattr(self.config, "model_file"):
-            if _is_master:
+            if _is_main:
                 warnings.warn(
                     "'model_file' key is required but missing "
                     "from FastTextProcessor's config."
@@ -442,7 +442,7 @@ def _try_download(self):
             model_file = os.path.join(get_mmf_cache_dir(), model_file)
 
         if not PathManager.exists(model_file):
-            if _is_master:
+            if _is_main:
                 warnings.warn(f"No model file present at {model_file}.")
             needs_download = True
 
@@ -455,11 +455,11 @@ def _try_download(self):
         synchronize()
 
     def _download_model(self):
-        _is_master = is_master()
+        _is_main = is_main()
 
         model_file_path = os.path.join(get_mmf_cache_dir(), "wiki.en.bin")
 
-        if not _is_master:
+        if not _is_main:
             return model_file_path
 
         if PathManager.exists(model_file_path):
@@ -477,7 +477,7 @@ def _download_model(self):
             pbar = tqdm(
                 total=int(response.headers["Content-Length"]) / 4096,
                 miniters=50,
-                disable=not _is_master,
+                disable=not _is_main,
             )
 
             idx = 0
@@ -726,7 +726,8 @@ class GraphVQAAnswerProcessor(BaseProcessor):
     "answers" or "answers_tokens". "answers" are preprocessed to generate
     "answers_tokens" if passed.
 
-    This version also takes a graph vocab and predicts a main and graph stream simultanously
+    This version also takes a graph vocab and predicts a main
+    and graph stream simultanously
 
     Args:
         config (DictConfig): Configuration for the processor

mmf/trainers/core/evaluation_loop.py

@@ -10,7 +10,7 @@
 from mmf.common.meter import Meter
 from mmf.common.report import Report
 from mmf.common.sample import to_device
-from mmf.utils.distributed import gather_tensor, is_master, is_xla
+from mmf.utils.distributed import gather_tensor, is_main, is_xla
 
 
 logger = logging.getLogger(__name__)
@@ -28,7 +28,7 @@ def evaluation_loop(
 
         with torch.no_grad():
             self.model.eval()
-            disable_tqdm = not use_tqdm or not is_master()
+            disable_tqdm = not use_tqdm or not is_main()
             while reporter.next_dataset(flush_report=False):
                 dataloader = reporter.get_dataloader()
                 combined_report = None

mmf/utils/build.py

@@ -18,7 +18,7 @@
 )
 from mmf.datasets.processors.processors import Processor
 from mmf.utils.configuration import Configuration, get_global_config
-from mmf.utils.distributed import is_dist_initialized, is_master, is_xla, synchronize
+from mmf.utils.distributed import is_dist_initialized, is_main, is_xla, synchronize
 from mmf.utils.general import get_optimizer_parameters
 from omegaconf import DictConfig, OmegaConf
 from packaging import version
@@ -96,7 +96,7 @@ def build_lightning_model(
     https://github.com/PyTorchLightning/pytorch-lightning/issues/5410
     """
 
-    if is_master():
+    if is_main():
         model_class.load_requirements(model_class, config=config)
         model = model_class.load_from_checkpoint(
             checkpoint_path, config=config, strict=False
@@ -140,7 +140,7 @@ def build_model(
         now other cores can proceed to build the model
         using already downloaded checkpoint.
         """
-        if is_master():
+        if is_main():
             model_class.load_requirements(model_class, config=config)
             model.build()
             synchronize()
@@ -250,7 +250,7 @@ def build_multiple_datamodules(
             )
             dataset_config = OmegaConf.create()
 
-        if is_master():
+        if is_main():
             datamodule_instance.prepare_data(dataset_config)
 
         synchronize()

mmf/utils/checkpoint.py

@@ -12,7 +12,7 @@
 from mmf.common.registry import registry
 from mmf.utils.checkpoint_updater import get_pretrained_state_mapping_checkpoint
 from mmf.utils.configuration import get_mmf_env, load_yaml
-from mmf.utils.distributed import is_master, is_xla, open_if_master, synchronize
+from mmf.utils.distributed import is_main, is_xla, open_if_main, synchronize
 from mmf.utils.download import download_pretrained_model
 from mmf.utils.file_io import PathManager
 from mmf.utils.general import get_current_device, updir
@@ -208,7 +208,7 @@ def __init__(self, trainer):
         self.saved_iterations = []
 
     def save_config(self):
-        if not is_master():
+        if not is_main():
             return
 
         cfg_file = os.path.join(self.ckpt_foldername, "config.yaml")
@@ -498,7 +498,7 @@ def save(self, update, iteration=None, update_best=False):
         # Which ensures that actual checkpoint saving happens
         # only for the master node.
         # The method also takes care of all the necessary synchronization
-        if not is_master() and not is_xla():
+        if not is_main() and not is_xla():
             return
 
         logger.info("Checkpoint save operation started!")
@@ -560,23 +560,23 @@ def save(self, update, iteration=None, update_best=False):
             git_metadata_dict = self._get_vcs_fields()
             ckpt.update(git_metadata_dict)
 
-        with open_if_master(ckpt_filepath, "wb") as f:
+        with open_if_main(ckpt_filepath, "wb") as f:
             self.save_func(ckpt, f)
 
         if update_best:
             logger.info("Saving best checkpoint")
-            with open_if_master(best_ckpt_filepath, "wb") as f:
+            with open_if_main(best_ckpt_filepath, "wb") as f:
                 self.save_func(ckpt, f)
 
         # Save current always
 
         logger.info("Saving current checkpoint")
-        with open_if_master(current_ckpt_filepath, "wb") as f:
+        with open_if_main(current_ckpt_filepath, "wb") as f:
             self.save_func(ckpt, f)
 
         # Remove old checkpoints if max_to_keep is set
         # In XLA, only delete checkpoint files in main process
-        if self.max_to_keep > 0 and is_master():
+        if self.max_to_keep > 0 and is_main():
             if len(self.saved_iterations) == self.max_to_keep:
                 self.remove(self.saved_iterations.pop(0))
             self.saved_iterations.append(update)
@@ -597,6 +597,6 @@ def restore(self):
             self._load(best_path, force=True)
 
     def finalize(self):
-        if is_master() or is_xla():
-            with open_if_master(self.pth_filepath, "wb") as f:
+        if is_main() or is_xla():
+            with open_if_main(self.pth_filepath, "wb") as f:
                 self.save_func(self.trainer.model.state_dict(), f)

mmf/utils/distributed.py

@@ -98,6 +98,10 @@ def get_rank():
     return dist.get_rank()
 
 
+def is_main():
+    return is_master()
+
+
 def is_master():
     return get_rank() == 0
 
@@ -380,20 +384,20 @@ def distributed_init(config):
         # perform a dummy all-reduce to initialize the NCCL communicator
         dist.all_reduce(torch.zeros(1).cuda())
 
-        suppress_output(is_master())
+        suppress_output(is_main())
         config.distributed.rank = dist.get_rank()
     return config.distributed.rank
 
 
-def suppress_output(is_master):
+def suppress_output(is_main):
     """Suppress printing on the current device. Force printing with `force=True`."""
     import builtins as __builtin__
 
     builtin_print = __builtin__.print
 
     def print(*args, **kwargs):
         force = kwargs.pop("force", False)
-        if is_master or force:
+        if is_main or force:
             builtin_print(*args, **kwargs)
 
     __builtin__.print = print
@@ -404,7 +408,7 @@ def print(*args, **kwargs):
 
     def warn(*args, **kwargs):
         force = kwargs.pop("force", False)
-        if is_master or force:
+        if is_main or force:
             builtin_warn(*args, **kwargs)
 
     # Log warnings only once
@@ -415,20 +419,24 @@ def warn(*args, **kwargs):
 def open_if_master(path, mode):
     from mmf.utils.file_io import PathManager
 
-    if is_master():
+    if is_main():
         return PathManager.open(path, mode)
     else:
         return contextlib.nullcontext()
 
 
+def open_if_main(*args):
+    return open_if_master(*args)
+
+
 def broadcast_xla_master_model_param(model):
     logger.info("Broadcasting XLA model parameters and buffers from master process ...")
 
     parameters_and_buffers = []
     for p in chain(model.parameters(), model.buffers()):
         # Set all params in non-master devices to zero so that all_reduce is equivalent
         # to broadcasting parameters from master to other devices.
-        if not is_master():
+        if not is_main():
             zero = torch.tensor(0, dtype=p.data.dtype, device=p.data.device)
             p.data.mul_(zero)
         parameters_and_buffers.append(p.data)

mmf/utils/early_stopping.py

@@ -1,7 +1,7 @@
 # Copyright (c) Facebook, Inc. and its affiliates.
 import numpy as np
 import torch
-from mmf.utils.distributed import is_master, is_xla
+from mmf.utils.distributed import is_main, is_xla
 
 
 class EarlyStopping:
@@ -49,7 +49,7 @@ def __call__(self, update, iteration, meter):
         # There are operations involving synchronization downstream
         # For XLA those calls must be executed from all cores
         # Therefore we do return here in case of XLA
-        if not is_master() and not is_xla():
+        if not is_main() and not is_xla():
             return False
 
         value = meter.meters.get(self.early_stop_criteria, None)

mmf/utils/logger.py

@@ -13,7 +13,7 @@
 import torch
 from mmf.common.registry import registry
 from mmf.utils.configuration import get_mmf_env
-from mmf.utils.distributed import get_rank, is_master, is_xla
+from mmf.utils.distributed import get_rank, is_main, is_xla
 from mmf.utils.file_io import PathManager
 from mmf.utils.timer import Timer
 from termcolor import colored
@@ -222,7 +222,7 @@ def summarize_report(
 ):
     if extra is None:
         extra = {}
-    if not is_master() and not is_xla():
+    if not is_main() and not is_xla():
         return
 
     if tb_writer:
@@ -309,15 +309,15 @@ def log_class_usage(component_type, klass):
 
 def skip_if_tensorboard_inactive(fn: Callable) -> Callable:
     """
-    Checks whether summary writer is initialized and rank is 0 (master)
+    Checks whether summary writer is initialized and rank is 0 (main)
     Args:
         fn (Callable): Function which should be called based on whether
             tensorboard should log or not
     """
 
     @wraps(fn)
     def wrapped_fn(self, *args: Any, **kwargs: Any) -> Optional[Any]:
-        if self.summary_writer is None or not self._is_master:
+        if self.summary_writer is None or not self._is_main:
             return None
         else:
             return fn(self, *args, **kwargs)
@@ -344,7 +344,7 @@ def formatMessage(self, record):
 class TensorboardLogger:
     def __init__(self, log_folder="./logs", iteration=0):
         self._summary_writer = None
-        self._is_master = is_master()
+        self._is_main = is_main()
         self.timer = Timer()
         self.log_folder = log_folder
         self.time_format = "%Y-%m-%dT%H:%M:%S"
@@ -356,7 +356,7 @@ def __init__(self, log_folder="./logs", iteration=0):
     @property
     def summary_writer(self):
         # Only on rank zero
-        if not self._is_master:
+        if not self._is_main:
             return None
 
         if self._summary_writer is None:
@@ -431,7 +431,7 @@ def setup(self):
         """
         Setup `Weights and Biases` for logging.
         """
-        if is_master():
+        if is_main():
 
             if self._wandb.run is None:
                 self._wandb.init(**self._wandb_init)
@@ -448,7 +448,7 @@ def __del__(self):
             self._wandb.finish()
 
     def _should_log_wandb(self):
-        if self._wandb is None or not is_master():
+        if self._wandb is None or not is_main():
             return False
         else:
             return True

mmf/utils/vocab.py

@@ -6,7 +6,7 @@
 import numpy as np
 import torch
 from mmf.utils.configuration import get_mmf_cache_dir
-from mmf.utils.distributed import is_master, synchronize
+from mmf.utils.distributed import is_main, synchronize
 from mmf.utils.file_io import PathManager
 from mmf.utils.general import get_absolute_path
 from torchtext import vocab
@@ -294,7 +294,7 @@ def __init__(self, vocab_file, embedding_name, *args, **kwargs):
 
         # First test loading the vectors in master so that everybody doesn't
         # download it in case it doesn't exist
-        if is_master():
+        if is_main():
             vocab.pretrained_aliases[embedding_name](cache=vector_cache)
         synchronize()
 
@@ -342,7 +342,7 @@ def __init__(self, embedding_name, *args, **kwargs):
 
         # First test loading the vectors in master so that everybody doesn't
         # download it in case it doesn't exist
-        if is_master():
+        if is_main():
             vocab.pretrained_aliases[embedding_name](cache=vector_cache)
         synchronize()
 

mmf/utils/xla.py

@@ -1,7 +1,7 @@
 # Copyright (c) Facebook, Inc. and its affiliates.
 
 import torch
-from mmf.utils.distributed import is_master
+from mmf.utils.distributed import is_main
 
 
 try:
@@ -16,10 +16,10 @@ def save_xla_ckpt(ckpt, file_or_path):
     checkpoint to CPU, since they hold PyTorch tensors. Other items like lr_scheduler
     often cannot be saved with xm.save due to its errors in handling mappingproxy.
 
-    Only save on the global master process (which is different from the default behavior
+    Only save on the global main process (which is different from the default behavior
     of xm.save that saves a checkpoint on each node).
     """
-    should_write_data = is_master()
+    should_write_data = is_main()
 
     is_full_ckpt = isinstance(ckpt, dict) and "model" in ckpt and "optimizer" in ckpt
     if is_full_ckpt:

projects/hateful_memes/README.md

@@ -14,7 +14,7 @@ Please cite the following paper if you use these models and the hateful memes da
   year={2020}
 }
 ```
-* [Citation for MMF](https://github.com/facebookresearch/mmf/tree/master/README.md#citation)
+* [Citation for MMF](https://github.com/facebookresearch/mmf/tree/main/README.md#citation)
 
 Links: [[arxiv]](https://arxiv.org/abs/2005.04790) [[challenge]](https://www.drivendata.org/competitions/70/hateful-memes-phase-2/) [[blog post]](https://ai.facebook.com/blog/hateful-memes-challenge-and-data-set)
 
@@ -54,7 +54,7 @@ In the table, we provide configuration corresponding to each of the baselines in
 | Visual BERT COCO | visual_bert    | visual_bert.finetuned.hateful_memes.from_coco    | projects/hateful_memes/configs/visual_bert/from_coco.yaml  |
 
 
-For individual baselines and their proper citation have a look at their project pages: [[Visual BERT]](https://github.com/facebookresearch/mmf/tree/master/projects/visual_bert) [[VilBERT]](https://github.com/facebookresearch/mmf/tree/master/projects/vilbert) [[MMBT]](https://github.com/facebookresearch/mmf/tree/master/projects/mmbt)
+For individual baselines and their proper citation have a look at their project pages: [[Visual BERT]](https://github.com/facebookresearch/mmf/tree/main/projects/visual_bert) [[VilBERT]](https://github.com/facebookresearch/mmf/tree/main/projects/vilbert) [[MMBT]](https://github.com/facebookresearch/mmf/tree/main/projects/mmbt)
 
 ## Training
 

projects/hateful_memes/fine_grained/README.md

@@ -6,13 +6,13 @@ This folder contains configs required to reproduce results and baselines for [20
 
 Install MMF following the [installation docs](https://mmf.sh/docs/getting_started/installation/).
 
-To acquire the hateful memes data, follow th instructions [here](https://github.com/facebookresearch/mmf/tree/master/projects/hateful_memes).
+To acquire the hateful memes data, follow th instructions [here](https://github.com/facebookresearch/mmf/tree/main/projects/hateful_memes).
 
-The additional fine grained labels can be found [here](https://github.com/facebookresearch/fine_grained_hateful_memes/tree/master/data).
+The additional fine grained labels can be found [here](https://github.com/facebookresearch/fine_grained_hateful_memes/tree/main/data).
 
 
 ## Reproducing Baselines
-We provide the configration fine to reproduce the baseline results we have in the [GitHub repo](https://github.com/facebookresearch/fine_grained_hateful_memes). The instrustions for training and evaluation are the same as [hateful memes](https://github.com/facebookresearch/mmf/tree/master/projects/hateful_memes). The output format is different and specified in the [GitHub repo](https://github.com/facebookresearch/fine_grained_hateful_memes).
+We provide the configration fine to reproduce the baseline results we have in the [GitHub repo](https://github.com/facebookresearch/fine_grained_hateful_memes). The instrustions for training and evaluation are the same as [hateful memes](https://github.com/facebookresearch/mmf/tree/main/projects/hateful_memes). The output format is different and specified in the [GitHub repo](https://github.com/facebookresearch/fine_grained_hateful_memes).
 
 The baselines are all based on VisualBert with image features.
 

website/docs/challenges/hateful_memes_challenge.md

@@ -6,13 +6,13 @@ sidebar_label: Hateful Memes Challenge
 
 The Hateful Memes challenge is available at [this link](https://www.drivendata.org/competitions/70/hateful-memes-phase-2/data/).
 
-In MMF, we provide the starter code and baseline pretrained models for this challenge and the configurations used for training the reported baselines. For more details check [this link](https://github.com/facebookresearch/mmf/tree/master/projects/hateful_memes).
+In MMF, we provide the starter code and baseline pretrained models for this challenge and the configurations used for training the reported baselines. For more details check [this link](https://github.com/facebookresearch/mmf/tree/main/projects/hateful_memes).
 
 In this tutorial, we provide steps for running training and evaluation with MMBT model on hateful memes dataset and generating submission file for the challenge. The same steps can be used for your own models.
 
 ## Installation and Preparing the dataset
 
-Follow the prerequisites for installation and dataset [here](https://github.com/facebookresearch/mmf/tree/master/projects/hateful_memes#prerequisites).
+Follow the prerequisites for installation and dataset [here](https://github.com/facebookresearch/mmf/tree/main/projects/hateful_memes#prerequisites).
 
 ## Training and Evaluation
 

website/docs/getting_started/installation.mdx

@@ -84,4 +84,4 @@ pytest ./tests/
 
 ## Contributing to MMF
 
-We welcome all contributions to MMF. Have a look at our [contributing guidelines](https://github.com/facebookresearch/mmf/tree/master/.github/CONTRIBUTING.md) to get started.
+We welcome all contributions to MMF. Have a look at our [contributing guidelines](https://github.com/facebookresearch/mmf/tree/main/.github/CONTRIBUTING.md) to get started.

website/docs/getting_started/quickstart.md

@@ -4,7 +4,7 @@ title: Quickstart
 sidebar_label: Quickstart
 ---
 
-In this quickstart guide, we are going to train the [M4C](https://github.com/facebookresearch/mmf/tree/master/projects/m4c) model on the TextVQA dataset. TextVQA requires models to read and reason about text in images to answer questions about them. `M4C` is a recent SOTA model on TextVQA which consists of a multimodal transformer architecture accompanied by a rich representation for text in images.
+In this quickstart guide, we are going to train the [M4C](https://github.com/facebookresearch/mmf/tree/main/projects/m4c) model on the TextVQA dataset. TextVQA requires models to read and reason about text in images to answer questions about them. `M4C` is a recent SOTA model on TextVQA which consists of a multimodal transformer architecture accompanied by a rich representation for text in images.
 
 To train other models or understand more about MMF, follow Next Steps at the bottom of this tutorial.
 

website/docs/notes/concepts.md

@@ -17,9 +17,9 @@ To achieve this, MMF has few opinions about architecture of your research projec
 
 ## Datasets
 
-You can find all the latest datasets [here](https://github.com/facebookresearch/mmf/tree/master/mmf/configs/datasets).
+You can find all the latest datasets [here](https://github.com/facebookresearch/mmf/tree/main/mmf/configs/datasets).
 
-The dataset's key is available under the particular dataset's config, ie., for vizwiz's key, you can look in vizwiz's config available [here](https://github.com/facebookresearch/mmf/blob/master/mmf/configs/datasets/vizwiz/defaults.yaml)
+The dataset's key is available under the particular dataset's config, ie., for vizwiz's key, you can look in vizwiz's config available [here](https://github.com/facebookresearch/mmf/blob/main/mmf/configs/datasets/vizwiz/defaults.yaml)
 
 ```yaml
 dataset_config:
@@ -41,11 +41,11 @@ Reference implementations for state-of-the-art models have been included to act
 - [Towards VQA Models That Can Read (LoRRA model)](https://arxiv.org/abs/1904.08920)
 - [VQA 2018 Challenge winner](https://arxiv.org/abs/1807.09956)
 - [VizWiz 2018 Challenge winner](https://vizwiz.org/wp-content/uploads/2019/06/workshop2018_slides_FAIR_A-STAR.pdf)
-- [VQA 2020 Challenge winner](https://github.com/facebookresearch/mmf/tree/master/projects/movie_mcan)
+- [VQA 2020 Challenge winner](https://github.com/facebookresearch/mmf/tree/main/projects/movie_mcan)
 
-Similar to datasets, each model has been registered with a unique key for easy reference in configuration and command line arguments. For a more complete list of models, please see [here](https://github.com/facebookresearch/mmf/tree/master/mmf/configs/models)
+Similar to datasets, each model has been registered with a unique key for easy reference in configuration and command line arguments. For a more complete list of models, please see [here](https://github.com/facebookresearch/mmf/tree/main/mmf/configs/models)
 
-The model's key is available under the particular model's config, ie., for mmf_transformer, the model's config file is available under [here](https://github.com/facebookresearch/mmf/blob/master/mmf/configs/models/mmf_transformer/defaults.yaml)
+The model's key is available under the particular model's config, ie., for mmf_transformer, the model's config file is available under [here](https://github.com/facebookresearch/mmf/blob/main/mmf/configs/models/mmf_transformer/defaults.yaml)
 
 ```yaml
 model_config:
@@ -76,11 +76,11 @@ Find more details about Registry class in its documentation [common/registry](ht
 
 ## Configuration
 
-As is necessary with research, most of the parameters/settings in MMF are configurable. MMF specific default values (`training`) are present in [mmf/configs/defaults.yaml](https://github.com/facebookresearch/mmf/blob/master/mmf/configs/defaults.yaml) with detailed comments delineating the usage of each parameter.
+As is necessary with research, most of the parameters/settings in MMF are configurable. MMF specific default values (`training`) are present in [mmf/configs/defaults.yaml](https://github.com/facebookresearch/mmf/blob/main/mmf/configs/defaults.yaml) with detailed comments delineating the usage of each parameter.
 
 For ease of usage and modularity, configuration for each dataset is kept separately in `mmf/configs/datasets/[dataset]/[variants].yaml` where you can get `[dataset]` value for the dataset from the tables in [Datasets](#datasets) section.
 
-The most dynamic part, model configurations are also kept separate and are the ones which need to be defined by the user if they are creating their own models. We include configurations for the models included in the model zoo of MMF. You can find the model configurations [here](https://github.com/facebookresearch/mmf/tree/master/mmf/configs/models)
+The most dynamic part, model configurations are also kept separate and are the ones which need to be defined by the user if they are creating their own models. We include configurations for the models included in the model zoo of MMF. You can find the model configurations [here](https://github.com/facebookresearch/mmf/tree/main/mmf/configs/models)
 
 
 It is possible to include other configs into your config using `includes` directive. Thus, in MMF config above you can include `lxmert`'s config like this:

website/docs/notes/configuration.md

@@ -9,7 +9,7 @@ MMF relies on [OmegaConf](https://omegaconf.readthedocs.io/en/latest/) for its c
 **TL;DR**
 
 - MMF uses OmegaConf for its configuration system with some sugar on top.
-- MMF defines [base defaults config](#base-defaults-config) containing all MMF specific parameters and then each dataset and model define their own configs (example configs: [[model]](https://github.com/facebookresearch/mmf/blob/master/mmf/configs/models/mmbt/defaults.yaml) [[dataset]](https://github.com/facebookresearch/mmf/blob/master/mmf/configs/datasets/hateful_memes/defaults.yaml)).
+- MMF defines [base defaults config](#base-defaults-config) containing all MMF specific parameters and then each dataset and model define their own configs (example configs: [[model]](https://github.com/facebookresearch/mmf/blob/main/mmf/configs/models/mmbt/defaults.yaml) [[dataset]](https://github.com/facebookresearch/mmf/blob/main/mmf/configs/datasets/hateful_memes/defaults.yaml)).
 - The user can define its own config specified by `config=<x>` at command line for each unique experiment or training setup. This has higher priority then base, model and dataset default configs and can override anything in those.
 - Finally, user can override (highest priority) the final config generated by merge of all above configs by specifying config parameters as [dotlist](https://omegaconf.readthedocs.io/en/latest/usage.html#from-a-dot-list) in their command. This is the **recommended** way of overriding the config parameters in MMF.
 - How MMF knows which config to pick for dataset and model? The user needs to specify those in his command as `model=x` and `dataset=y`.
@@ -100,7 +100,7 @@ model_config:
 
 ## User Config
 
-User can specify their configuration specific to an experiment or training setup by adding `config=<config_path>` argument to their command. User config can specify for e.g. training parameters according to their experiment such as batch size using `training.batch_size`. Most common use case for user config is to specify optimizer, scheduler and training parameters. Other than that user config can also include configs for variations of models and datasets they want to test on. Have a look at an example user config [here](https://github.com/facebookresearch/mmf/blob/master/projects/hateful_memes/configs/mmbt/defaults.yaml).
+User can specify their configuration specific to an experiment or training setup by adding `config=<config_path>` argument to their command. User config can specify for e.g. training parameters according to their experiment such as batch size using `training.batch_size`. Most common use case for user config is to specify optimizer, scheduler and training parameters. Other than that user config can also include configs for variations of models and datasets they want to test on. Have a look at an example user config [here](https://github.com/facebookresearch/mmf/blob/main/projects/hateful_memes/configs/mmbt/defaults.yaml).
 
 ## Command Line Dot List Override
 
@@ -210,4 +210,4 @@ MMF supports overriding some of the config parameters through environment variab
 
 ## Base Defaults Config
 
-Have a look at the [defaults config of MMF](https://github.com/facebookresearch/mmf/blob/master/mmf/configs/defaults.yaml) along with description of parameters from which you may need to override parameters for your experiments.
+Have a look at the [defaults config of MMF](https://github.com/facebookresearch/mmf/blob/main/mmf/configs/defaults.yaml) along with description of parameters from which you may need to override parameters for your experiments.

website/docs/notes/projects.md

@@ -6,13 +6,13 @@ sidebar_label: MMF Projects
 
 MMF contains references implementations or has been used to develop following projects (in no particular order):
 
-- Iterative Answer Prediction with Pointer-Augmented Multimodal Transformers for TextVQA [[arXiv](https://arxiv.org/abs/1911.06258)] [[project](https://github.com/facebookresearch/mmf/tree/master/projects/m4c)]
-- ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks [[arXiv](https://arxiv.org/abs/1908.02265)] [[project](https://github.com/facebookresearch/mmf/tree/master/projects/vilbert)]
+- Iterative Answer Prediction with Pointer-Augmented Multimodal Transformers for TextVQA [[arXiv](https://arxiv.org/abs/1911.06258)] [[project](https://github.com/facebookresearch/mmf/tree/main/projects/m4c)]
+- ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks [[arXiv](https://arxiv.org/abs/1908.02265)] [[project](https://github.com/facebookresearch/mmf/tree/main/projects/vilbert)]
 - Visualbert: A simple and performant baseline for vision and language [[arXiv](https://arxiv.org/abs/1908.03557)] [[project](https://arxiv.org/abs/1908.03557)]
-- The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes [[arXiv](https://arxiv.org/abs/2005.04790)] [[project](https://github.com/facebookresearch/mmf/tree/master/projects/hateful_memes)]
-- Towards VQA Models That Can Read [[arXiv](https://arxiv.org/abs/1904.08920)] [[project](https://github.com/facebookresearch/mmf/tree/master/projects/lorra)]
-- TextCaps: a Dataset for Image Captioning with Reading Comprehension [[arXiv](https://arxiv.org/abs/2003.12462)] [[project](https://github.com/facebookresearch/mmf/tree/master/projects/m4c_captioner)]
-- Pythia v0. 1: the winning entry to the vqa challenge 2018 [[arXiv](https://arxiv.org/abs/1807.09956)] [[project](https://github.com/facebookresearch/mmf/tree/master/projects/pythia)]
-- Bottom-up and top-down attention for image captioning and visual question answering [[arXiv](https://arxiv.org/abs/1707.07998)] [[project](https://github.com/facebookresearch/mmf/tree/master/projects/butd)]
-- Supervised Multimodal Bitransformers for Classifying Images and Text [[arXiv](https://arxiv.org/abs/1909.02950)] [[project](https://github.com/facebookresearch/mmf/tree/master/projects/mmbt)]
-- Are we pretraining it right? Digging deeper into visio-linguistic pretraining [[arXiv](https://arxiv.org/abs/2004.08744)][[project](https://github.com/facebookresearch/mmf/tree/master/projects/pretrain_vl_right)]
+- The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes [[arXiv](https://arxiv.org/abs/2005.04790)] [[project](https://github.com/facebookresearch/mmf/tree/main/projects/hateful_memes)]
+- Towards VQA Models That Can Read [[arXiv](https://arxiv.org/abs/1904.08920)] [[project](https://github.com/facebookresearch/mmf/tree/main/projects/lorra)]
+- TextCaps: a Dataset for Image Captioning with Reading Comprehension [[arXiv](https://arxiv.org/abs/2003.12462)] [[project](https://github.com/facebookresearch/mmf/tree/main/projects/m4c_captioner)]
+- Pythia v0. 1: the winning entry to the vqa challenge 2018 [[arXiv](https://arxiv.org/abs/1807.09956)] [[project](https://github.com/facebookresearch/mmf/tree/main/projects/pythia)]
+- Bottom-up and top-down attention for image captioning and visual question answering [[arXiv](https://arxiv.org/abs/1707.07998)] [[project](https://github.com/facebookresearch/mmf/tree/main/projects/butd)]
+- Supervised Multimodal Bitransformers for Classifying Images and Text [[arXiv](https://arxiv.org/abs/1909.02950)] [[project](https://github.com/facebookresearch/mmf/tree/main/projects/mmbt)]
+- Are we pretraining it right? Digging deeper into visio-linguistic pretraining [[arXiv](https://arxiv.org/abs/2004.08744)][[project](https://github.com/facebookresearch/mmf/tree/main/projects/pretrain_vl_right)]

website/docs/projects/m4c.md

@@ -35,7 +35,7 @@ The released imdbs contain OCR results and normalized bounding boxes (i.e. in th
 
 For the TextVQA dataset, the downloaded file contains both imdbs with the Rosetta-en OCRs (better performance) and imdbs with Rosetta-ml OCRs (same OCR results as in the previous [LoRRA](http://openaccess.thecvf.com/content_CVPR_2019/papers/Singh_Towards_VQA_Models_That_Can_Read_CVPR_2019_paper.pdf) model). Please download the corresponding OCR feature files.
 
-Note that the object Faster R-CNN features are extracted with [`extract_features_vmb.py`](https://github.com/facebookresearch/mmf/blob/master/tools/scripts/features/extract_features_vmb.py) and the OCR Faster R-CNN features are extracted with [`extract_ocr_frcn_feature.py`](https://github.com/facebookresearch/mmf/blob/master/projects/m4c/scripts/extract_ocr_frcn_feature.py).
+Note that the object Faster R-CNN features are extracted with [`extract_features_vmb.py`](https://github.com/facebookresearch/mmf/blob/main/tools/scripts/features/extract_features_vmb.py) and the OCR Faster R-CNN features are extracted with [`extract_ocr_frcn_feature.py`](https://github.com/facebookresearch/mmf/blob/main/projects/m4c/scripts/extract_ocr_frcn_feature.py).
 
 ## Pretrained M4C Models
 

website/docs/tutorials/checkpointing.md

@@ -8,7 +8,7 @@ In this tutorial, we will learn about the different details around finetuning fr
 
 ## Pre-requisites and installation
 
-Follow the prerequisites for installation and dataset [here](https://github.com/facebookresearch/mmf/tree/master/projects/hateful_memes#prerequisites).
+Follow the prerequisites for installation and dataset [here](https://github.com/facebookresearch/mmf/tree/main/projects/hateful_memes#prerequisites).
 
 ## Finetuning from a pretrained model
 

website/docs/tutorials/concat_bert_tutorial.md

@@ -4,13 +4,13 @@ title: 'Tutorial: Adding a model - Concat BERT'
 sidebar_label: Adding a model - Concat BERT
 ---
 
-In this tutorial, we will go through the step-by-step process of creating a new model using MMF. In this case, we will create a fusion model and train it on the [Hateful Memes dataset](https://github.com/facebookresearch/mmf/tree/master/projects/hateful_memes).
+In this tutorial, we will go through the step-by-step process of creating a new model using MMF. In this case, we will create a fusion model and train it on the [Hateful Memes dataset](https://github.com/facebookresearch/mmf/tree/main/projects/hateful_memes).
 
 The fusion model that we will create concatenates embeddings from a text encoder and an image encoder and passes them through a two-layer classifier. MMF provides standard image and text encoders out of the box. For the image encoder, we will use ResNet152 image encoder and for the text encoder, we will use BERT-Base Encoder.
 
 ## Prerequisites
 
-Follow the prerequisites for installation and dataset [here](https://github.com/facebookresearch/mmf/tree/master/projects/hateful_memes#prerequisites).
+Follow the prerequisites for installation and dataset [here](https://github.com/facebookresearch/mmf/tree/main/projects/hateful_memes#prerequisites).
 
 ## Using MMF to build the model
 

website/docs/tutorials/losses.md

@@ -10,8 +10,8 @@ This is a tutorial on how to add a new loss function to MMF.
 
 MMF is agnostic to the kind of losses that can be added to it.
 Adding a loss requires adding a loss class and adding your new loss to your config yaml.
-For example, the [ConcatBERT](https://github.com/facebookresearch/mmf/blob/master/website/docs/tutorials/concat_bert_tutorial.md) model uses the `cross_entropy` loss when training on the hateful memes dataset.
-The loss class is `CrossEntropyLoss` defined in [mmf/modules/losses.py](https://github.com/facebookresearch/mmf/blob/master/mmf/modules/losses.py)
+For example, the [ConcatBERT](https://github.com/facebookresearch/mmf/blob/main/website/docs/tutorials/concat_bert_tutorial.md) model uses the `cross_entropy` loss when training on the hateful memes dataset.
+The loss class is `CrossEntropyLoss` defined in [mmf/modules/losses.py](https://github.com/facebookresearch/mmf/blob/main/mmf/modules/losses.py)
 The loss key `cross_entropy` is added to the list of losses in the config yaml at [mmf/projects/hateful_memes/configs/concat_bert/defaults.yaml](https://github.com/facebookresearch/mmf/blob/15fa63071bfaed56db43deba871cfec76439c66f/projects/others/concat_bert/hateful_memes/defaults.yaml#L11).
 
 
@@ -62,7 +62,7 @@ For losses with params you can do,
 
 If a loss class is responsible for calculating multiple losses, for example, maybe due to shared calculations you can return a dictionary of tensors.
 The resulting loss that is optimized is the sum of all losses configured for the model.
-For an example, take a look at the `BCEAndKLLoss` class in [mmf/modules/losses.py](https://github.com/facebookresearch/mmf/blob/master/mmf/modules/losses.py)
+For an example, take a look at the `BCEAndKLLoss` class in [mmf/modules/losses.py](https://github.com/facebookresearch/mmf/blob/main/mmf/modules/losses.py)
 
 ```python
 @registry.register_loss("bce_kl")

website/docs/tutorials/metrics.md

@@ -10,8 +10,8 @@ This is a tutorial on how to add a new metric to MMF.
 
 MMF is agnostic to the kind of metrics that can be added to it.
 Adding a metric requires adding a metric class and adding your new metric to your config yaml.
-For example, the [ConcatBERT](https://github.com/facebookresearch/mmf/blob/master/website/docs/tutorials/concat_bert_tutorial.md) model uses the `binary_f1` metric when evaluating on the hateful memes dataset.
-The metric class is `BinaryF1` defined in [mmf/modules/metrics.py](https://github.com/facebookresearch/mmf/blob/master/mmf/modules/metrics.py)
+For example, the [ConcatBERT](https://github.com/facebookresearch/mmf/blob/main/website/docs/tutorials/concat_bert_tutorial.md) model uses the `binary_f1` metric when evaluating on the hateful memes dataset.
+The metric class is `BinaryF1` defined in [mmf/modules/metrics.py](https://github.com/facebookresearch/mmf/blob/main/mmf/modules/metrics.py)
 The metric key `binary_f1` is added to the list of metrics in the config yaml at [mmf/projects/hateful_memes/configs/concat_bert/defaults.yaml](https://github.com/facebookresearch/mmf/blob/15fa63071bfaed56db43deba871cfec76439c66f/projects/others/concat_bert/hateful_memes/defaults.yaml#L28).
 
 
@@ -88,7 +88,7 @@ evaluation:
   - binary_f1
 ```
 
-For metrics that take parameters your yaml config will specify params. You can also specify a custom key to be assigned to the metric. For [example](https://github.com/facebookresearch/mmf/blob/master/projects/unit/configs/vg/single_task.yaml),
+For metrics that take parameters your yaml config will specify params. You can also specify a custom key to be assigned to the metric. For [example](https://github.com/facebookresearch/mmf/blob/main/projects/unit/configs/vg/single_task.yaml),
 
 ```yaml
 evaluation:
@@ -104,7 +104,7 @@ evaluation:
 
 ```
 
-If your model uses early stopping, make sure that the early_stop.criteria is added as an evaluation metric. For example the [vizwiz](https://github.com/facebookresearch/mmf/blob/master/projects/ban/configs/vizwiz/defaults.yaml) config,
+If your model uses early stopping, make sure that the early_stop.criteria is added as an evaluation metric. For example the [vizwiz](https://github.com/facebookresearch/mmf/blob/main/projects/ban/configs/vizwiz/defaults.yaml) config,
 
 ```yaml
 evaluation:
@@ -121,7 +121,7 @@ training:
 
 If a loss class is responsible for calculating multiple metrics, for example, maybe due to shared calculations, you can return a dictionary of tensors.
 
-For an example, take a look at the `BinaryF1PrecisionRecall` class in [mmf/modules/metrics.py](https://github.com/facebookresearch/mmf/blob/master/mmf/modules/metrics.py)
+For an example, take a look at the `BinaryF1PrecisionRecall` class in [mmf/modules/metrics.py](https://github.com/facebookresearch/mmf/blob/main/mmf/modules/metrics.py)
 
 ```python
 @registry.register_metric("f1_precision_recall")

website/docs/tutorials/slurm.md

@@ -82,7 +82,7 @@ Add `--dry_run` argument to first print out what exactly is going to be run with
 
 :::
 
-An actual complex sweep config for visual bert with more options can be found at [./tools/sweep/sweep_visual_bert.py](https://github.com/facebookresearch/mmf/blob/master/tools/sweeps/sweep_visual_bert.py). Command following the above command to run it:
+An actual complex sweep config for visual bert with more options can be found at [./tools/sweep/sweep_visual_bert.py](https://github.com/facebookresearch/mmf/blob/main/tools/sweeps/sweep_visual_bert.py). Command following the above command to run it:
 
 ```sh
 python tools/sweeps/sweep_visual_bert.py \

website/docusaurus.config.js

@@ -111,7 +111,7 @@ module.exports = {
           showLastUpdateTime: true,
           editUrl: fbContent({
             internal: 'https://www.internalfb.com/intern/diffusion/FBS/browse/master/fbcode/faim/mmf/website',
-            external: 'https://github.com/facebookresearch/mmf/edit/master/website/'
+            external: 'https://github.com/facebookresearch/mmf/edit/main/website/'
           }),
         },
         theme: {