6.mdx

<FrameworkSwitchCourse {fw} />

End-of-chapter quiz[[end-of-chapter-quiz]]

Let's test what you learned in this chapter!

1. What are models on the Hub limited to?

<Question choices={[ { text: "Models from the 🤗 Transformers library.", explain: "While models from the 🤗 Transformers library are supported on the Hugging Face Hub, they're not the only ones!" }, { text: "All models with a similar interface to 🤗 Transformers.", explain: "No interface requirement is set when uploading models to the Hugging Face Hub. " }, { text: "There are no limits.", explain: "Right! There are no limits when uploading models to the Hub.", correct: true }, { text: "Models that are in some way related to NLP.", explain: "No requirement is set regarding the field of application!" } ]} />

2. How can you manage models on the Hub?

<Question choices={[ { text: "Through a GCP account.", explain: "Incorrect!" }, { text: "Through peer-to-peer distribution.", explain: "Incorrect!" }, { text: "Through git and git-lfs.", explain: "Correct! Models on the Hub are simple Git repositories, leveraging git-lfs for large files.", correct: true } ]} />

3. What can you do using the Hugging Face Hub web interface?

<Question choices={[ { text: "Fork an existing repository.", explain: "Forking a repository is not possible on the Hugging Face Hub." }, { text: "Create a new model repository.", explain: "Correct! That's not all you can do, though.", correct: true }, { text: "Manage and edit files.", explain: "Correct! That's not the only right answer, though.", correct: true }, { text: "Upload files.", explain: "Right! But that's not all.", correct: true }, { text: "See diffs across versions.", explain: "Correct! That's not all you can do, though.", correct: true } ]} />

4. What is a model card?

<Question choices={[ { text: "A rough description of the model, therefore less important than the model and tokenizer files.", explain: "It is indeed a description of the model, but it's an important piece: if it's incomplete or absent the model's utility is drastically reduced." }, { text: "A way to ensure reproducibility, reusability, and fairness.", explain: "Correct! Sharing the right information in the model card will help users leverage your model and be aware of its limits and biases. ", correct: true }, { text: "A Python file that can be run to retrieve information about the model.", explain: "Model cards are simple Markdown files." } ]} />

5. Which of these objects of the 🤗 Transformers library can be directly shared on the Hub with push_to_hub()?

{#if fw === 'pt'} <Question choices={[ { text: "A tokenizer", explain: "Correct! All tokenizers have the push_to_hub method, and using it will push all the tokenizer files (vocabulary, architecture of the tokenizer, etc.) to a given repo. That's not the only right answer, though!", correct: true }, { text: "A model configuration", explain: "Right! All model configurations have the push_to_hub method, and using it will push them to a given repo. What else can you share?", correct: true }, { text: "A model", explain: "Correct! All models have the push_to_hub method, and using it will push them and their configuration files to a given repo. That's not all you can share, though.", correct: true }, { text: "A Trainer", explain: "That's right — the Trainer also implements the push_to_hub method, and using it will upload the model, its configuration, the tokenizer, and a model card draft to a given repo. Try another answer!", correct: true } ]} /> {:else} <Question choices={[ { text: "A tokenizer", explain: "Correct! All tokenizers have the push_to_hub method, and using it will push all the tokenizer files (vocabulary, architecture of the tokenizer, etc.) to a given repo. That's not the only right answer, though!", correct: true }, { text: "A model configuration", explain: "Right! All model configurations have the push_to_hub method, and using it will push them to a given repo. What else can you share?", correct: true }, { text: "A model", explain: "Correct! All models have the push_to_hub method, and using it will push them and their configuration files to a given repo. That's not all you can share, though.", correct: true }, { text: "All of the above with a dedicated callback", explain: "That's right — the PushToHubCallback will regularly send all of those objects to a repo during training.", correct: true } ]} /> {/if}

6. What is the first step when using the push_to_hub() method or the CLI tools?

<Question choices={[ { text: "Log in on the website.", explain: "This won't help you on your local machine." }, { text: "Run 'huggingface-cli login' in a terminal.", explain: "Correct — this will download and cache your personal token.", correct: true }, { text: "Run 'notebook_login()' in a notebook.", explain: "Correct — this will display a widget to let you authenticate.", correct: true }, ]} />

7. You're using a model and a tokenizer — how can you upload them to the Hub?

<Question choices={[ { text: "By calling the push_to_hub method directly on the model and the tokenizer.", explain: "Correct!", correct: true }, { text: "Within the Python runtime, by wrapping them in a huggingface_hub utility.", explain: "Models and tokenizers already benefit from huggingface_hub utilities: no need for additional wrapping!" }, { text: "By saving them to disk and calling transformers-cli upload-model", explain: "The command upload-model does not exist." } ]} />

8. Which git operations can you do with the Repository class?

<Question choices={[ { text: "A commit.", explain: "Correct, the git_commit() method is there for that.", correct: true }, { text: "A pull", explain: "That is the purpose of the git_pull() method.", correct: true }, { text: "A push", explain: "The method git_push() does this.", correct: true }, { text: "A merge", explain: "No, that operation will never be possible with this API." } ]} />