[2.0] Reodering arguments for torch jit #1010 and future TF2.0 compatibility #1195
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
thomwolf
force-pushed
the
reorder_arguments
branch
from
16296da to
7fba47b
Compare
Codecov Report
@@ Coverage Diff @@
## master #1195 +/- ##
==========================================
- Coverage 80.83% 80.42% -0.41%
==========================================
Files 46 46
Lines 7878 7892 +14
==========================================
- Hits 6368 6347 -21
- Misses 1510 1545 +35
Continue to review full report at Codecov.
|
thomwolf
changed the title
LysandreJik
reviewed
Sep 5, 2019
| outputs = self.bert(input_ids, position_ids=position_ids, token_type_ids=token_type_ids, | ||
| attention_mask=attention_mask, head_mask=head_mask) | ||
| def forward(self, input_ids, attention_mask=None, token_type_ids=None, | ||
| position_ids=None, head_mask=None, labels=None): |
LysandreJik
reviewed
Sep 5, 2019
| if attention_mask is not None: | ||
| # Apply the attention mask | ||
| w = w + attention_mask | ||
|
|
LysandreJik
reviewed
Sep 5, 2019
|
|
||
| if attention_mask is not None: | ||
| # Apply the attention mask | ||
| w = w + attention_mask |
LysandreJik
reviewed
Sep 5, 2019
| model.eval() | ||
| sequence_output, pooled_output = model(input_ids, token_type_ids, input_mask) | ||
| sequence_output, pooled_output = model(input_ids, token_type_ids) | ||
| sequence_output, pooled_output = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids) |
Member
There was a problem hiding this comment.
I feel like this is clearer than before! Looks great.
LysandreJik
approved these changes
Sep 5, 2019
Member
LysandreJik
left a comment
LysandreJik
left a comment
There was a problem hiding this comment.
This seems to solve the problems related to JIT. I feel it's a great addition as the order of inputs now accurately depicts the order of importance.
Member
Author
|
Ok merging |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Torch jit (cf #1010) and TF 2.0 (cf #1104) are more strict than PyTorch on having a specific order of arguments for easy use.
This PR refactor the order of the keyword arguments to make them as natural as possible.
This will be a breaking change for people using positional order to input keyword arguments in the forward pass of the models, hence is delayed to the 2.0 release.