Transformer divisibility error validation #3105
Merged
Conversation
This file contains 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
justinxzhao
reviewed
Feb 16, 2023
ludwig/config_validation/checks.py
Outdated
Comment on lines
353
to
360
| sequence_types = [SEQUENCE, TEXT, TIMESERIES] | ||
| for sequence_type in sequence_types: | ||
| encoder = config.defaults.__getattribute__(sequence_type).encoder | ||
| if encoder.type == "transformer" and not is_divisible(encoder.hidden_size, encoder.num_heads): | ||
| raise ConfigValidationError( | ||
| f"Default {sequence_type} transformer encoder requires encoder.hidden_size to be divisible by " | ||
| f"encoder.num_heads. Found hidden_size {encoder.hidden_size} and num_heads {encoder.num_heads}." | ||
| ) |
There was a problem hiding this comment.
I think we might be able to get away without checking config.defaults as all of these auxiliary checks are run after defaults have been resolved into feature configs. WDYT?
There was a problem hiding this comment.
Oh, cool! I'll remove the defaults block.
| } | ||
|
|
||
| with pytest.raises(ConfigValidationError): | ||
| validate_config(config) |
There was a problem hiding this comment.
Use ModelConfig.from_dict(config) instead of validate_config, which I'm planning on removing in #3104
There was a problem hiding this comment.
So all of the tests fail when I use ModelConfig.from_dict. Do we need to wait for the PR to land before updating?
There was a problem hiding this comment.
@jeffkinnison you're totally right! Let me try to land that PR first.
jeffkinnison
force-pushed
the
transformer-divisibility-error
branch
from
81d0be2 to
e325b86
Compare
justinxzhao
approved these changes
Feb 17, 2023
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.
Followup to #3066 that adds a validation check for
encoder.hidden_size % encoder.num_heads == 0for transformer encoders.