[Fix] fix bugs when loading KV files and more. Also make horovod saving function is callable independent bug not patched origin save function. #364
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
MoFHeka
changed the title
fuhailin
reviewed
Oct 23, 2023
| @@ -233,8 +233,7 @@ def _insert_de_shard_from_file_system( | |||
| _values_tensor_dataset = readers.FixedLengthRecordDataset( | |||
| shard_values_file_list, | |||
| record_bytes=_value_dtype.size * value_dim, | |||
| buffer_size=buffer_size * value_dim).padded_batch(insert_num_once * | |||
| value_dim, | |||
There was a problem hiding this comment.
Why here need to delete value_dim?
There was a problem hiding this comment.
_values_tensor_dataset record_bytes is already _value_dtype.size * value_dim, if we padded_batch the bytes, they could be the size _value_dtype.size * value_dim * insert_num_once * value_dim. When they are reshaped, their shape would not compatible with _keys_tensor_dataset.
MoFHeka
changed the title
MoFHeka
force-pushed
the
master-dev
branch
3 times, most recently
from
5a4dfb4
to
f4404dc
Compare
…set op, which could cause insert error.
… this time, as this can lead to unexpected errors. Now using de.keras.models.de_hvd_save_model to replace tf.keras.models.save_model.
…n, the recovered data will be the same first data in vector.
…estore parameter when using tf.train.Saver.
…nd is used to save DE KV files with different rank when using horovod all2all training.
…ger when using HvdAllToAllEmbedding by calling de.train.DEHvdCheckpoint.
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Outdated suggestions cannot be applied.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.
Description
Functional Fix
Fix the default save function for tensorflow will not be patched at this time, as this can lead to unexpected errors.
Now using de.keras.models.de_hvd_save_model to replace tf.keras.models.save_model.
Bug Fix
New Feature
New Document
[doc] Add documentation for HvdAllToAllEmbedding.
Type of change
Checklist:
How Has This Been Tested?
Run the test.
But the fatal error in CPU table only happened in template <class K, class V, size_t DIM>, and I don't know how to trigger it in a small case.