HuggingFace --> Hugging Face

README.md

@@ -26,7 +26,7 @@ The library provides:
 - Support for checkpoints in various formats, including checkpoints in HF format
 - Training recipes for popular fine-tuning techniques with reference benchmarks and comprehensive correctness checks
 - Evaluation of trained models with EleutherAI Eval Harness
-- Integration with HuggingFace Datasets for training
+- Integration with Hugging Face Datasets for training
 - Support for distributed training using FSDP from PyTorch Distributed
 - YAML configs for easily configuring training runs
 - [Upcoming] Support for lower precision dtypes and quantization techniques from [TorchAO](https://github.com/pytorch-labs/ao)
@@ -182,7 +182,7 @@ TorchTune embodies PyTorch’s design philosophy [[details](https://pytorch.org/
 
 #### Native PyTorch
 
-TorchTune is a native-PyTorch library. While we provide integrations with the surrounding ecosystem (eg: HuggingFace Datasets, EluetherAI Eval Harness), all of the core functionality is written in PyTorch.
+TorchTune is a native-PyTorch library. While we provide integrations with the surrounding ecosystem (eg: Hugging Face Datasets, EluetherAI Eval Harness), all of the core functionality is written in PyTorch.
 
 #### Simplicity and Extensibility
 

docs/source/examples/first_finetune_tutorial.rst

@@ -25,13 +25,13 @@ job using TorchTune.
 Downloading a model
 -------------------
 First, you need to download a model. TorchTune's supports an integration
-with the `HuggingFace Hub <https://huggingface.co/docs/hub/en/index>`_ - a collection of the latest and greatest model weights.
+with the `Hugging Face Hub <https://huggingface.co/docs/hub/en/index>`_ - a collection of the latest and greatest model weights.
 
 For this tutorial, you're going to use the `Llama2 model from Meta <https://llama.meta.com/>`_. Llama2 is a "gated model",
 meaning that you need to be granted access in order to download the weights. Follow `these instructions <https://huggingface.co/meta-llama>`_ on the official Meta page
-hosted on HuggingFace to complete this process. (This should take less than 5 minutes.)
+hosted on Hugging Face to complete this process. (This should take less than 5 minutes.)
 
-Once you have authorization, you will need to authenticate with HuggingFace Hub. The easiest way to do so is to provide an
+Once you have authorization, you will need to authenticate with Hugging Face Hub. The easiest way to do so is to provide an
 access token to the download script. You can find your token `here <https://huggingface.co/settings/tokens>`_.
 
 Then, it's as simple as:

docs/source/overview.rst

@@ -22,7 +22,7 @@ TorchTune provides:
 - Modular native-PyTorch implementations of popular LLMs
 - Interoperability with popular model zoos through checkpoint-conversion utilities
 - Training recipes for a variety of fine-tuning techniques
-- Integration with `HuggingFace Datasets <https://huggingface.co/docs/datasets/en/index>`_ for training and `EleutherAI's Eval <https://github.com/EleutherAI/lm-evaluation-harness>`_ Harness for evaluation
+- Integration with `Hugging Face Datasets <https://huggingface.co/docs/datasets/en/index>`_ for training and `EleutherAI's Eval <https://github.com/EleutherAI/lm-evaluation-harness>`_ Harness for evaluation
 - Support for distributed training using `FSDP <https://pytorch.org/docs/stable/fsdp.html>`_
 - Yaml configs for easily configuring training runs
 
@@ -55,7 +55,7 @@ TorchTune embodies `PyTorch’s design philosophy <https://pytorch.org/docs/stab
 
 **Native PyTorch**
 
-TorchTune is a native-PyTorch library. While we provide integrations with the surrounding ecosystem (eg: HuggingFace Datasets, EluetherAI Eval Harness), all of the core functionality is written in PyTorch.
+TorchTune is a native-PyTorch library. While we provide integrations with the surrounding ecosystem (eg: Hugging Face Datasets, EluetherAI Eval Harness), all of the core functionality is written in PyTorch.
 
 
 **Simplicity and Extensibility**

requirements.txt

@@ -1,4 +1,4 @@
-# HuggingFace Integration Reqs
+# Hugging Face Integration Reqs
 datasets
 huggingface_hub
 

tests/torchtune/_cli/test_download.py

@@ -45,7 +45,7 @@ def test_download_calls_snapshot(self, capsys, monkeypatch, snapshot_download):
         with pytest.raises(SystemExit, match="2"):
             runpy.run_path(TUNE_PATH, run_name="__main__")
         err = capsys.readouterr().err
-        assert "not found on the HuggingFace Hub" in err
+        assert "not found on the Hugging Face Hub" in err
 
         # Call the third time and get the expected output
         runpy.run_path(TUNE_PATH, run_name="__main__")

tests/torchtune/data/test_templates.py

@@ -14,7 +14,7 @@
     SummarizeTemplate,
 )
 
-# Taken from Open-Orca/SlimOrca-Dedup on HuggingFace:
+# Taken from Open-Orca/SlimOrca-Dedup on Hugging Face:
 # https://huggingface.co/datasets/Open-Orca/SlimOrca-Dedup
 CHAT_SAMPLE = {
     "system": "You are an AI assistant. User will you give you a task. Your goal is to complete the task as faithfully as you can. While performing the task think step-by-step and justify your steps.",  # noqa: B950

torchtune/_cli/download.py

@@ -24,28 +24,28 @@ def __init__(self, subparsers: argparse._SubParsersAction):
             "download",
             prog="tune download",
             usage="tune download <repo-id> [OPTIONS]",
-            help="Download a model from the HuggingFace Hub.",
-            description="Download a model from the HuggingFace Hub.",
+            help="Download a model from the Hugging Face Hub.",
+            description="Download a model from the Hugging Face Hub.",
             epilog=textwrap.dedent(
                 """\
             examples:
-                # Download a model from the HuggingFace Hub with a Hugging Face API token
+                # Download a model from the Hugging Face Hub with a Hugging Face API token
                 $ tune download meta-llama/Llama-2-7b-hf --hf-token <TOKEN> --output-dir /tmp/model
                 Successfully downloaded model repo and wrote to the following locations:
                 ./model/config.json
                 ./model/README.md
                 ./model/consolidated.00.pth
                 ...
 
-                # Download an ungated model from the HuggingFace Hub
+                # Download an ungated model from the Hugging Face Hub
                 $ tune download mistralai/Mistral-7B-Instruct-v0.2
                 Successfully downloaded model repo and wrote to the following locations:
                 ./model/config.json
                 ./model/README.md
                 ./model/model-00001-of-00002.bin
                 ...
 
-            For a list of all models, visit the HuggingFace Hub https://huggingface.co/models.
+            For a list of all models, visit the Hugging Face Hub https://huggingface.co/models.
             """
             ),
             formatter_class=argparse.RawTextHelpFormatter,
@@ -58,7 +58,7 @@ def _add_arguments(self) -> None:
         self._parser.add_argument(
             "repo_id",
             type=str,
-            help="Name of the repository on HuggingFace Hub.",
+            help="Name of the repository on Hugging Face Hub.",
         )
         self._parser.add_argument(
             "--output-dir",
@@ -72,7 +72,7 @@ def _add_arguments(self) -> None:
             type=str,
             required=False,
             default=os.getenv("HF_TOKEN", None),
-            help="HuggingFace API token. Needed for gated models like Llama2.",
+            help="Hugging Face API token. Needed for gated models like Llama2.",
         )
         self._parser.add_argument(
             "--ignore-patterns",
@@ -84,7 +84,7 @@ def _add_arguments(self) -> None:
         )
 
     def _download_cmd(self, args: argparse.Namespace) -> None:
-        """Downloads a model from the HuggingFace Hub."""
+        """Downloads a model from the Hugging Face Hub."""
         # Download the tokenizer and PyTorch model files
         try:
             true_output_dir = snapshot_download(
@@ -96,13 +96,13 @@ def _download_cmd(self, args: argparse.Namespace) -> None:
         except GatedRepoError:
             self._parser.error(
                 "It looks like you are trying to access a gated repository. Please ensure you "
-                "have access to the repository and have provided the proper HuggingFace API token "
+                "have access to the repository and have provided the proper Hugging Face API token "
                 "using the option `--hf-token` or by running `huggingface-cli login`."
                 "You can find your token by visiting https://huggingface.co/settings/tokens"
             )
         except RepositoryNotFoundError:
             self._parser.error(
-                f"Repository '{args.repo_id}' not found on the HuggingFace Hub."
+                f"Repository '{args.repo_id}' not found on the Hugging Face Hub."
             )
         except Exception as e:
             self._parser.error(e)

torchtune/data/_templates.py

@@ -240,7 +240,7 @@ class ChatMLTemplate(PromptTemplate):
     """
     OpenAI's Chat Markup Language used by their chat models:
     https://github.com/MicrosoftDocs/azure-docs/blob/main/articles/ai-services/openai/includes/chat-markup-language.md
-    It is the default template used by HuggingFace models.
+    It is the default template used by Hugging Face models.
 
     Example:
         <|im_start|>system

torchtune/datasets/_alpaca.py

@@ -15,7 +15,7 @@ def alpaca_dataset(
     use_clean: bool = False,
 ) -> InstructDataset:
     """
-    Support for the Alpaca dataset and its variants from HuggingFace Datasets.
+    Support for the Alpaca dataset and its variants from Hugging Face Datasets.
     https://huggingface.co/datasets/tatsu-lab/alpaca
 
     Data input format: https://huggingface.co/datasets/tatsu-lab/alpaca#data-instances

torchtune/datasets/_chat.py

@@ -45,7 +45,7 @@ class ChatDataset(Dataset):
 
     Args:
         tokenizer (Tokenizer): Tokenizer used to encode data. Tokenize must implement an `encode` and `decode` method.
-        source (str): path string of dataset, anything supported by HuggingFace's `load_dataset`
+        source (str): path string of dataset, anything supported by Hugging Face's `load_dataset`
             (https://huggingface.co/docs/datasets/en/package_reference/loading_methods#datasets.load_dataset.path)
         convert_to_dialogue (Callable[[Mapping[str, Any]], Dialogue]): function that keys into the desired field in the sample
             and converts to a list of `Messages` that follows the llama format with the expected keys
@@ -151,7 +151,7 @@ def chat_dataset(
 
     Args:
         tokenizer (Tokenizer): Tokenizer used to encode data. Tokenize must implement an `encode` and `decode` method.
-        source (str): path string of dataset, anything supported by HuggingFace's `load_dataset`
+        source (str): path string of dataset, anything supported by Hugging Face's `load_dataset`
             (https://huggingface.co/docs/datasets/en/package_reference/loading_methods#datasets.load_dataset.path)
         conversation_format (str): string specifying expected format of conversations in the dataset
             for automatic conversion to the llama format. Supported formats are: "sharegpt"

torchtune/datasets/_grammar.py

@@ -14,7 +14,7 @@ def grammar_dataset(
     train_on_input: bool = False,
 ) -> InstructDataset:
     """
-    Support for the Grammar dataset and its variants from HuggingFace Datasets.
+    Support for the Grammar dataset and its variants from Hugging Face Datasets.
     https://huggingface.co/datasets/liweili/c4_200m
 
     Data input format: https://huggingface.co/datasets/liweili/c4_200m#description

torchtune/datasets/_instruct.py

@@ -34,7 +34,7 @@ class InstructDataset(Dataset):
 
     Args:
         tokenizer (Tokenizer): Tokenizer used to encode data. Tokenize must implement an `encode` and `decode` method.
-        source (str): path string of dataset, anything supported by HuggingFace's `load_dataset`
+        source (str): path string of dataset, anything supported by Hugging Face's `load_dataset`
             (https://huggingface.co/docs/datasets/en/package_reference/loading_methods#datasets.load_dataset.path)
         template (PromptTemplate): template used to format the prompt. If the placeholder variable
             names in the template do not match the column/key names in the dataset, use `column_map` to map them.
@@ -103,7 +103,7 @@ def instruct_dataset(
 
     Args:
         tokenizer (Tokenizer): Tokenizer used to encode data. Tokenize must implement an `encode` and `decode` method.
-        source (str): path string of dataset, anything supported by HuggingFace's `load_dataset`
+        source (str): path string of dataset, anything supported by Hugging Face's `load_dataset`
             (https://huggingface.co/docs/datasets/en/package_reference/loading_methods#datasets.load_dataset.path)
         template (str): class name of template used to format the prompt. If the placeholder variable
             names in the template do not match the column/key names in the dataset, use `column_map` to map them.

torchtune/datasets/_samsum.py

@@ -14,7 +14,7 @@ def samsum_dataset(
     train_on_input: bool = False,
 ) -> InstructDataset:
     """
-    Support for the Summarize dataset and its variants from HuggingFace Datasets.
+    Support for the Summarize dataset and its variants from Hugging Face Datasets.
     https://huggingface.co/datasets/samsum
 
     Data input format: https://huggingface.co/datasets/samsum#data-fields

torchtune/modules/lr_schedulers.py

@@ -22,7 +22,7 @@ def get_cosine_schedule_with_warmup(
     0.0 to lr over num_warmup_steps, then decreases to 0.0 on a cosine schedule over
     the remaining num_training_steps-num_warmup_steps (assuming num_cycles = 0.5).
 
-    This is based on the HuggingFace implementation
+    This is based on the Hugging Face implementation
     https://github.com/huggingface/transformers/blob/v4.23.1/src/transformers/optimization.py#L104.
 
     Args: