Explicit arguments in from_pretrained
#24306
Conversation
|
The documentation is not available anymore as the PR was closed or merged. |
There was a problem hiding this comment.
Thanks for owrking on this. I think ignore_mismatched_sizes and use_safetensors might be worth exposing as well. We should also rework the docstring to group the arguments exposed first and the the others second (with titles like main arguments and power user arguments).
|
TODO:
|
|
@sgugger Would be nice if you can take a quick look 🙏 . And do you want me to deal with all framework (TF/Flax), tokenizer/processor, and also |
ydshieh
changed the title
from_pretrainedfrom_pretrained
|
You have a lot of tests failing to fix 😅 , sure you want a review yet? |
|
@sgugger No, I didn't request a new review since last time you have a look. But the changes pushed triggered you 😆 |
| def _set_token_in_kwargs(self, kwargs, token=None): | ||
| """Temporary method to deal with `token` and `use_auth_token`. | ||
|
|
||
| This method is to avoid apply the same changes in all model config classes that overwrite `from_pretrained`. | ||
|
|
||
| Need to clean up `use_auth_token` in a follow PR. | ||
| """ | ||
| # Some model config classes like CLIP define their own `from_pretrained` without the new argument `token` yet. | ||
| if token is None: | ||
| token = kwargs.pop("token", None) | ||
| use_auth_token = kwargs.pop("use_auth_token", None) | ||
|
|
||
| if use_auth_token is not None: | ||
| warnings.warn( | ||
| "The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers.", FutureWarning | ||
| ) | ||
| if token is not None: | ||
| raise ValueError( | ||
| "`token` and `use_auth_token` are both specified. Please set only the argument `token`." | ||
| ) | ||
| token = use_auth_token | ||
|
|
||
| if token is not None: | ||
| # change to `token` in a follow-up PR | ||
| kwargs["use_auth_token"] = token |
There was a problem hiding this comment.
We have quite some model classes (e.g. clip-like family) whose config classes have their own from_pretrained.
This private _set_token_in_kwargs method is to make life easier when dealing with token and use_auth_token.
We will see what the best way is in a follow up PR when we want to make those customized from_pretrained with explicit arguments.
There was a problem hiding this comment.
@sgugger Now it's the good time if you are motivated after 🦷.
There was a problem hiding this comment.
Ugh... Annoying! This fix works in the meantime.
| def _set_token_in_kwargs(self, kwargs, token=None): | ||
| """Temporary method to deal with `token` and `use_auth_token`. | ||
|
|
||
| This method is to avoid apply the same changes in all model config classes that overwrite `from_pretrained`. | ||
|
|
||
| Need to clean up `use_auth_token` in a follow PR. | ||
| """ | ||
| # Some model config classes like CLIP define their own `from_pretrained` without the new argument `token` yet. | ||
| if token is None: | ||
| token = kwargs.pop("token", None) | ||
| use_auth_token = kwargs.pop("use_auth_token", None) | ||
|
|
||
| if use_auth_token is not None: | ||
| warnings.warn( | ||
| "The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers.", FutureWarning | ||
| ) | ||
| if token is not None: | ||
| raise ValueError( | ||
| "`token` and `use_auth_token` are both specified. Please set only the argument `token`." | ||
| ) | ||
| token = use_auth_token | ||
|
|
||
| if token is not None: | ||
| # change to `token` in a follow-up PR | ||
| kwargs["use_auth_token"] = token |
There was a problem hiding this comment.
Ugh... Annoying! This fix works in the meantime.
What does this PR do?
[still incomplete]
Need to apply the same changes to other files containing
from_pretrained(other framework, other classes like config, processor, auto, etc.) but @sgugger let me know if I am not lost already in the early stage.