Fix typos and add Typo check GitHub Action

[shirayu]

• Fix typos
• Add typo check GitHub Action: https://github.com/crate-ci/typos

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[anton-l]

Pretty cool addition IMO, I'm open to try it in manual mode for now, and then implement nightly/weekly typo-fixing PRs if everything goes well :) Thanks @shirayu!
@patrickvonplaten @patil-suraj @pcuenca wdyt?

[anton-l]

Look like this is what was meant in those docstrings:

Suggested change

	nd="np" # nd may be np (numpy)
	nd.array="np.ndarray"

[pcuenca]

I don't understand how this works, from the documentation it looks like it's used to add words to the dictionary (to ignore them), or is it to force corrections? Or both? How was the nd.array typo detected in the first place, was it Typo that pointed it out or did @shirayu saw it themselves and so decided to add it here?

[shirayu]

I tried typo detection of np.ndarra with this change but it did not work.
My understanding is that typos first tokenize the file and then check using the dictionary.
Therefore, the words that contain . are not recognized well.

[anton-l]

Could you link a line where this appears please?

[shirayu]

Hereis the result of typos without the line nin="nin".
https://gist.githubusercontent.com/shirayu/43649cc2a3b32c875c40027c02b0c946/raw/44e2be398f73f72cb5971113754c2e4b85e7c177/NIN_at_diffusers

At first I thought nin was an abbreviation for something, but is it a typo for min?

[anton-l]

Looking at https://github.com/huggingface/diffusers/blame/5dda1735fda047f4242d28f91e6e457b9760d52d/src/diffusers/models/resnet.py#L325
I think use_in_shortcut was meant here @patrickvonplaten?

[shirayu]

Thank you. I fixed use_in_shortcut at b75558b.

-        new_item = new_item.replace("nin_shortcut", "conv_shortcut")
+        new_item = new_item.replace("in_shortcut", "conv_shortcut")

Please check whether this change is OK.

[shirayu]

Another use of MIN is below.

error: `NIN` should be `INN`, `MIN`, `BIN`, `NINE`
  --> ./scripts/convert_ncsnpp_original_checkpoint_to_diffusers.py:49:76
   |
49 |         new_layer.query.weight.data = old_checkpoint[f"all_modules.{index}.NIN_0.W"].data.T
   |                                                                            ^^^
   |
error: `NIN` should be `INN`, `MIN`, `BIN`, `NINE`
  --> ./scripts/convert_ncsnpp_original_checkpoint_to_diffusers.py:50:74
   |
50 |         new_layer.key.weight.data = old_checkpoint[f"all_modules.{index}.NIN_1.W"].data.T
   |                                                                          ^^^
   |
error: `NIN` should be `INN`, `MIN`, `BIN`, `NINE`
  --> ./scripts/convert_ncsnpp_original_checkpoint_to_diffusers.py:51:76
   |
51 |         new_layer.value.weight.data = old_checkpoint[f"all_modules.{index}.NIN_2.W"].data.T
   |                                                                            ^^^
   |
error: `NIN` should be `INN`, `MIN`, `BIN`, `NINE`
  --> ./scripts/convert_ncsnpp_original_checkpoint_to_diffusers.py:53:74
   |
53 |         new_layer.query.bias.data = old_checkpoint[f"all_modules.{index}.NIN_0.b"].data
   |                                                                          ^^^
   |
error: `NIN` should be `INN`, `MIN`, `BIN`, `NINE`
  --> ./scripts/convert_ncsnpp_original_checkpoint_to_diffusers.py:54:72
   |
54 |         new_layer.key.bias.data = old_checkpoint[f"all_modules.{index}.NIN_1.b"].data
   |                                                                        ^^^
   |
error: `NIN` should be `INN`, `MIN`, `BIN`, `NINE`
  --> ./scripts/convert_ncsnpp_original_checkpoint_to_diffusers.py:55:74
   |
55 |         new_layer.value.bias.data = old_checkpoint[f"all_modules.{index}.NIN_2.b"].data
   |                                                                          ^^^
   |
error: `NIN` should be `INN`, `MIN`, `BIN`, `NINE`
  --> ./scripts/convert_ncsnpp_original_checkpoint_to_diffusers.py:57:80
   |
57 |         new_layer.proj_attn.weight.data = old_checkpoint[f"all_modules.{index}.NIN_3.W"].data.T
   |                                                                                ^^^
   |
error: `NIN` should be `INN`, `MIN`, `BIN`, `NINE`
  --> ./scripts/convert_ncsnpp_original_checkpoint_to_diffusers.py:58:78
   |
58 |         new_layer.proj_attn.bias.data = old_checkpoint[f"all_modules.{index}.NIN_3.b"].data
   |                                                                              ^^^
   |

I think this should not be changed.
I added a memo about this at b7845a3.

[pcuenca]

It's surprising the number of typos that were detected!

What does the action do when it runs, does it open a PR like this one, or what would the correction workflow look like? (I'm thinking it could just be a part of the Makefile instead, maybe?)

[pcuenca]

I don't understand how this works, from the documentation it looks like it's used to add words to the dictionary (to ignore them), or is it to force corrections? Or both? How was the nd.array typo detected in the first place, was it Typo that pointed it out or did @shirayu saw it themselves and so decided to add it here?

[patrickvonplaten]

Very cool! I'd be happy to add it this one since it should be cheap. But we need indeed to be careful that no "correct" words, definitions are accidently changed

[patil-suraj]

Very cool! As @anton-l said we could try this in manual mode to start with :)

[patrickvonplaten]

Good for me if you think it's useful @anton-l :-)

[patrickvonplaten]

Happy to follow @shirayu's and @anton-l's lead here

[anton-l]

Ok, let's do a trial run then! :)