Use Value dim shape for Attention compute_output_shape #19284
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
fchollet
reviewed
Mar 11, 2024
| def test_attention_compute_output_shape(self): | ||
| layer = layers.Attention() | ||
| input_shape = [(2, 8, 7), (2, 8, 5), (2, 8, 7)] # Shapes of Q, V, K | ||
| self.assertAllEqual(layer.compute_output_shape(input_shape) == (2, 8, 5)) |
There was a problem hiding this comment.
Please call the layer on an input and read its shape, to ensure match between actual shape and computed shape.
There was a problem hiding this comment.
Updated to match it with output.shape
sampathweb
force-pushed
the
use-value-dim-shape-attention
branch
from
9b26297
to
ececf6e
Compare
sampathweb
commented
Mar 18, 2024
| @@ -173,9 +173,6 @@ def _parse_name(self, name): | |||
| return "float16", "float32" | |||
| elif name == "mixed_bfloat16": | |||
| return "bfloat16", "float32" | |||
| elif name == "uint8": | |||
There was a problem hiding this comment.
This is redundant. Its addressed in the try block so removing it.
google-ml-butler
bot
added
kokoro:force-run
ready to pull
google-ml-butler
bot
removed
ready to pull
james77777778
added a commit
to james77777778/keras
that referenced
this pull request
Mar 19, 2024
…nse` Add qlora-like technique to `quantized_call` in `Dense` Update `save_own_variables` and `load_own_variables` Update `benchmark.py` update version string. Set dtype policy for uint8 (keras-team#19327) * Set Quantization policy for uint8 to float * Add uint8 to dtype_policies Use Value dim shape for Attention compute_output_shape (keras-team#19284) * Use Value dim shape for Attention compute_output_shape * Fix attention layer compute output shape * fix format * check compute_output_shape with output Update `quantized_call` in `EinsumDense` to support training with quantized weights
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 #19257 by using Value dim shape for
compute_output_shapeof Attention layer.