This change moves the initialization of commonly used `SymbolicExpr` and a sample `SymbolicMap` into the `SymbolicMapTest` fixture to reduce code duplication across tests.
PiperOrigin-RevId: 826161168
I originally assumed the caller was always providing a full list of replacements but IndexingMap have some uses where the dim_replacement list is empty, resulting in a CHECK-fail.
So, I'm allowing the user to provide either dim or symbol empty lists to ReplaceDimsAndSymbols. In that case, the dims/symbols won't be replaced.
PiperOrigin-RevId: 826138814
The GlobalToLocal and LocalToGlobal custom-calls are for Shardy round trip. These get-tuple-elements will be removed when we import the Shardy dialect and thus they do not need to hold frontend attributes.
This can reduce the size of the generated HLO module text.
PiperOrigin-RevId: 826134489
Instead of 2 ra2a + concat, we can double the output buffer and adjust output offsets. This way we can save on latency by having only one multi-GPU synchronization.
PiperOrigin-RevId: 826122665
- The logic is ported from gemm_fusion_autotuner. I have changed the key of the Key Value store to be just module-fingerprint, earlier it was module-fingerprint + autotunable-fusion-set-from-the-module-fingerprint. The module fingerprint should already represent the fusion-sets contained in it.
- We can improve or just remove this functionality when we design storage for offline autotuning.
PiperOrigin-RevId: 826103885
This prevents max chains of trivial ops (e.g. add.fp32) to run faster than copying the data, which results in 'too fast to measure' error.
PiperOrigin-RevId: 826079017
This change updates the Bazel version used in TensorFlow, JAX, and XLA projects from 7.4.1 to 7.7.0 in `.bazelversion` files and build scripts.
PiperOrigin-RevId: 826075658
Imported from GitHub PR https://github.com/openxla/xla/pull/33278
Bumps [keras](https://github.com/keras-team/keras) from 3.11.3 to 3.12.0.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/keras-team/keras/releases">keras's releases</a>.</em></p>
<blockquote>
<h2>Keras 3.12.0</h2>
<h2>Highlights</h2>
<h3>Keras has a new model distillation API!</h3>
<p>You now have access to an easy-to-use API for distilling large models into small models while minimizing performance drop on a reference dataset -- compatible with all existing Keras models. You can specify a range of different distillation losses, or create your own losses. The API supports multiple concurrent distillation losses at the same time.</p>
<p>Example:</p>
<pre lang="python"><code># Load a model to distill
teacher = ...
# This is the model we want to distill it into
student = ...
<h1>Configure the process</h1>
<p>distiller = Distiller(
teacher=teacher,
student=student,
distillation_losses=LogitsDistillation(temperature=3.0),
)
distiller.compile(
optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy']
)</p>
<h1>Train the distilled model</h1>
<p>distiller.fit(x_train, y_train, epochs=10)
</code></pre></p>
<h3>Keras supports GPTQ quantization!</h3>
<p>GPTQ is now built into the Keras API. GPTQ is a post-training, weights-only quantization method that compresses a model to int4 layer by layer. For each layer, it uses a second-order method to update weights while minimizing the error on a calibration dataset.</p>
<p>Learn how to use it <a href="https://keras.io/guides/gptq_quantization_in_keras/">in this guide</a>.</p>
<p>Example:</p>
<pre lang="python"><code>model = keras_hub.models.Gemma3CausalLM.from_preset("gemma3_1b")
gptq_config = keras.quantizers.GPTQConfig(
dataset=calibration_dataset,
tokenizer=model.preprocessor.tokenizer,
weight_bits=4,
group_size=128,
num_samples=256,
sequence_length=256,
hessian_damping=0.01,
symmetric=False,
</tr></table>
</code></pre>
</blockquote>
<p>... (truncated)</p>
</details>
<details>
<summary>Commits</summary>
<ul>
<li><a href="adbfd13426"><code>adbfd13</code></a> Add warning to <code>set_backend</code> and more detailed example. (<a href="https://redirect.github.com/keras-team/keras/issues/21787">#21787</a>)</li>
<li><a href="70598b7903"><code>70598b7</code></a> Fix typo in Distiller docstring</li>
<li><a href="eecd34f406"><code>eecd34f</code></a> Fix: <code>keras.ops.quantile</code> works with tf graph execution (<a href="https://redirect.github.com/keras-team/keras/issues/21782">#21782</a>)</li>
<li><a href="c2bc6cfcc7"><code>c2bc6cf</code></a> Suport keras.op.view() to view the same data bitwise at a new dtype (<a href="https://redirect.github.com/keras-team/keras/issues/21763">#21763</a>)</li>
<li><a href="10b51ce5a5"><code>10b51ce</code></a> Make confusion metrics compilable. (<a href="https://redirect.github.com/keras-team/keras/issues/21775">#21775</a>)</li>
<li><a href="18f79d69c9"><code>18f79d6</code></a> Fix negative index handling in MultiHeadAttention attention_axes (<a href="https://redirect.github.com/keras-team/keras/issues/21721">#21721</a>)</li>
<li><a href="18e0364cbc"><code>18e0364</code></a> Support for extracting volume patches (<a href="https://redirect.github.com/keras-team/keras/issues/21759">#21759</a>)</li>
<li><a href="dc5e42cca4"><code>dc5e42c</code></a> fix sas metrics in jax <code>fit</code> (<a href="https://redirect.github.com/keras-team/keras/issues/21765">#21765</a>)</li>
<li><a href="1ba3b8f896"><code>1ba3b8f</code></a> Fix discretization discrepancy (<a href="https://redirect.github.com/keras-team/keras/issues/21769">#21769</a>)</li>
<li><a href="53987a768d"><code>53987a7</code></a> Document that <code>set_backend</code> requires re-importing keras. (<a href="https://redirect.github.com/keras-team/keras/issues/21764">#21764</a>)</li>
<li>Additional commits viewable in <a href="https://github.com/keras-team/keras/compare/v3.11.3...v3.12.0">compare view</a></li>
</ul>
</details>
<br />
[](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores)
Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it
- `@dependabot merge` will merge this PR after your CI passes on it
- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it
- `@dependabot cancel merge` will cancel a previously requested merge and block automerging
- `@dependabot reopen` will reopen this PR if it is closed
- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually
- `@dependabot show <dependency name> ignore conditions` will show all of the ignore conditions of the specified dependency
- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)
You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/openxla/xla/network/alerts).
</details>
Copybara import of the project:
--
b37d94a32428d62ed3e73765f4e7b61bc6ed8549 by dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>:
Bump keras in /xla/backends/cpu/benchmarks/e2e/gemma2/keras
Bumps [keras](https://github.com/keras-team/keras) from 3.11.3 to 3.12.0.
- [Release notes](https://github.com/keras-team/keras/releases)
- [Commits](https://github.com/keras-team/keras/compare/v3.11.3...v3.12.0)
---
updated-dependencies:
- dependency-name: keras
dependency-version: 3.12.0
dependency-type: direct:production
...
Signed-off-by: dependabot[bot] <support@github.com>
Merging this change closes#33278
PiperOrigin-RevId: 826053656
Symbolic dimensions and symbols are both implemented as SymbolicExpr variables, with symbols being offset by the number of dimensions. This implementation detail was previously exposed to users of SymbolicExprContext, who had to manually calculate variable IDs. I ended having a hidden bug in the implementation of IndexingMap, so I included these free functions to make the translation less bug-prone.
The SymbolicMap tests have been updated to use these new helper functions
PiperOrigin-RevId: 826053249
Imported from GitHub PR https://github.com/google-ai-edge/LiteRT/pull/4272
Summary:
- Add typename to avoid cmake failure
Test Result:
======================== Test Summary ========================
//litert/vendors/qualcomm/core/utils:utils_test
[==========] 12 tests from 3 test suites ran. (0 ms total)
[ PASSED ] 12 tests.
YOU HAVE 2 DISABLED TESTS
//litert/vendors/qualcomm/core/backends:qnn_backend_test
[==========] 0 tests from 0 test suites ran. (0 ms total)
[ PASSED ] 0 tests.
YOU HAVE 4 DISABLED TESTS
//litert/vendors/qualcomm/core/wrappers/tests:op_wrapper_test
[==========] 7 tests from 1 test suite ran. (0 ms total)
[ PASSED ] 7 tests.
//litert/vendors/qualcomm/core/wrappers/tests:tensor_wrapper_test
[==========] 18 tests from 1 test suite ran. (0 ms total)
[ PASSED ] 18 tests.
//litert/vendors/qualcomm/core/wrappers/tests:param_wrapper_test
[==========] 16 tests from 2 test suites ran. (0 ms total)
[ PASSED ] 16 tests.
//litert/vendors/qualcomm/core/wrappers/tests:quantize_params_wrapper_test
[==========] 13 tests from 3 test suites ran. (0 ms total)
[ PASSED ] 13 tests.
//litert/vendors/qualcomm/core:common_test
[==========] 13 tests from 1 test suite ran. (0 ms total)
[ PASSED ] 13 tests.
//litert/vendors/qualcomm/core:tensor_pool_test
[==========] 8 tests from 1 test suite ran. (0 ms total)
[ PASSED ] 8 tests.
//litert/vendors/qualcomm:qnn_manager_test
[==========] 3 tests from 1 test suite ran. (259 ms total)
[ PASSED ] 3 tests.
//litert/c/options:litert_qualcomm_options_test
[==========] 17 tests from 2 test suites ran. (0 ms total)
[ PASSED ] 17 tests.
//litert/c:litert_op_options_test
//litert/tools/flags/vendors:qualcomm_flags_test
[==========] 8 tests from 5 test suites ran. (0 ms total)
[ PASSED ] 8 tests.
//litert/vendors/qualcomm/compiler:qnn_compiler_plugin_test
[==========] 232 tests from 4 test suites ran. (61193 ms total)
[ PASSED ] 232 tests.
Copybara import of the project:
--
57e7429920d1950d635b29c4f8a18ccd51ade496 by chuntl-qti <chuntl@qti.qualcomm.com>:
Qualcomm AI Engine Direct - Fix android-arm64 CMake build bug
Summary:
- Add typename to avoid cmake failure
Merging this change closes#4272
PiperOrigin-RevId: 826045177
Add CustomCallThunk::OwnedHandlerBundle, a bag of `unique_ptr<ffi::Ffi>` that
enable using lambdas with captures in CustomCallThunk. Lambda captures must
outlive the created thunk.
The functionality is similar to what is possible with "old-style" callbacks,
but doesn't depend on them, and adds support for other handlers available via
XLA_FFI_Handler_Bundle.
PiperOrigin-RevId: 826043689
This change implements AbslHashValue and llvm::hash_value for xla::gpu::SymbolicMap.
This is a prerequisite for correctly implementing AbslHashValue for xla::IndexingMap after its internal migration to use SymbolicMap. Specifically, it needs be used in IndexingMap::AbslHashValue.
PiperOrigin-RevId: 826038011
`ffi::Attribute` and related types are members of the `CustomCallThunk`. Therefore we need to be able to serialize these types to proto message if we wanna be able to serialize instances of CustomCallThunk.
So this change is adding proto message representation of the types and adds functions `ToProto` and `FromProto` to each of them. Most of the types are currently defined as type aliases of some `std::variant` instantiation. This changes replaces the aliases by classes which inherit from the std::variant type. These new types then get the proto serialization functions.
PiperOrigin-RevId: 826035931
