Pcs#41
Open
nathanchenseanwalter wants to merge 16 commits into
Open
Conversation
…s://github.com/anchal-physics/keras2c Thanks for providing this.
Change doc links from f0uriest to plasmacontrol
nathanchenseanwalter
force-pushed
the
pcs
branch
from
dc8cf68 to
c506e1f
Compare
This is reflected in the C implementation k2c_activations.c This resolved test_LSTM1, test_ProfilePredictorConv2D Made average pooling more robust. There was a zero division with improper padding. This is resolved and solves test_AveragePooling1D2, test_AveragePooling2D2 Keras3 changed the name 'swish' to 'silu'. This fixes the test test_swish The graph construction of bidirectional layers changed from Keras2 to Keras3 and led to extra layers, hence, numerical instabilities. This is now resolved. Also, smaller fixes in test_split_layers.py.
…s were not resolved properly. This is now fixed and tests included in the test suite.
Added one more test for activation after a dense layer.
* Algorithmic C optimizations + batch norm folding - affine_matmul: i-k-j loop reorder with bias init fusion - batch_norm: precomputed scale/bias, eliminates integer division - relu/ReLU: branchless ternary for vectorization - Code generator: fold batch_norm into adjacent Dense/Conv1D layers - Fix: BatchNorm negative axis handling for Keras 3 * Add __restrict and __attribute__((hot)) annotations (no flag changes) * Add static const codegen + fix ReLU codegen bug * Remove unneeded files. * Add optimization tests and deep cloning. --------- Co-authored-by: Matthew Waller <hello@cephalopod.studio>
* Algorithmic C optimizations + batch norm folding - affine_matmul: i-k-j loop reorder with bias init fusion - batch_norm: precomputed scale/bias, eliminates integer division - relu/ReLU: branchless ternary for vectorization - Code generator: fold batch_norm into adjacent Dense/Conv1D layers - Fix: BatchNorm negative axis handling for Keras 3 * Add __restrict and __attribute__((hot)) annotations (no flag changes) * Add static const codegen + fix ReLU codegen bug * Remove unneeded files. * Add optimization tests and deep cloning. * Add defensive callouts for threading. --------- Co-authored-by: Matthew Waller <hello@cephalopod.studio>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
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.
Update critical components and up-to-date documentation