Fix Mistral memory consumption with JAX and default dtype bug #1460
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
mattdangerw
approved these changes
Feb 23, 2024
| @@ -300,7 +300,7 @@ def main(_): | |||
| print("-> Saved the model weights in float16") | |||
|
|
|||
| # === Save the model config === | |||
| keras_nlp_config["dtype"] = "bfloat16" | |||
| keras_nlp_config.pop("dtype") # We don't want a default dtype | |||
There was a problem hiding this comment.
I would actually rename keras_nlp_config -> backbone_kwargs, and save using
keras_nlp.src.utils.preset_utils.save_to_preset(
keras_nlp_model, preset
)
keras_nlp.src.utils.preset_utils.save_to_preset(
keras_nlp_tokenizer, preset, config_filename="tokenizer.json"
)
Those will do the right thing and always call model.get_config() to create the config.json file (which does not include dtype for this reason). No need to regenerate presets if things look good in testing.
There was a problem hiding this comment.
Done, uploaded the new presets to Kaggle.
mattdangerw
pushed a commit
that referenced
this pull request
Feb 27, 2024
abuelnasr0
pushed a commit
to abuelnasr0/keras-nlp
that referenced
this pull request
Apr 2, 2024
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Outdated suggestions cannot be applied.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.
Fixes #1458
This PR updates the presets for the Mistral model. The configs have been updated on Kaggle to not set a default dtype. The JAX memory consumption bug should also be fixed now.