https://gitcode.net/xusiwei1236/tflite-micro/-/commit/d2c4ef56f7a67db2da97f42de1f67e404f13eaadAutomated sync from github.com/tensorflow/tensorflow (#2080)2023-06-26T07:52:25+00:00TFLM-bottflm-github-bot@google.com
BUG=automated sync from upstream
NO_CHECK_TFLITE_FILES=automated sync from upstreamhttps://gitcode.net/xusiwei1236/tflite-micro/-/commit/360f05d2943252e02b02642612d42c2885ca5158Signal Library OPs python BUILD update (#2088)2023-06-27T20:25:59+00:00suleshahid110432064+suleshahid@users.noreply.github.com
This is needed to properly build all the ops together. Since we are calling into a singular utils function to load the ops, if it checks for any other ops, it would break before this.
Now we make it build all ops everytime utils.py is built.
BUG=[288938993](<a href="http://b/288938993" rel="nofollow noreferrer noopener" target="_blank">http://b/288938993</a>)https://gitcode.net/xusiwei1236/tflite-micro/-/commit/5d2b26d85ea176d0b735ae39f5dacf762e55e97eAutomated sync from github.com/tensorflow/tensorflow (#2087)2023-06-27T21:48:58+00:00TFLM-bottflm-github-bot@google.com
BUG=automated sync from upstream
NO_CHECK_TFLITE_FILES=automated sync from upstreamhttps://gitcode.net/xusiwei1236/tflite-micro/-/commit/d8b44a5f5f4b3bc0e6138f2cbf57d2055f2b4ea1IF kernel Eval not using TfLiteEvalTensor (#2086)2023-06-28T00:01:32+00:00David Davisddavis-2015@users.noreply.github.com
@tensorflow/micro
The IF kernel has a non-compliant implementation. Eval phase is not using TfLiteEvalTensor.
The `cond_value` variable should be declared `const`
bug=fixes #2043https://gitcode.net/xusiwei1236/tflite-micro/-/commit/8ace118b04cc0e50a88a3ed45ef53516a95ec0d1Update Signal RFFT to have namespace and reformat Makefile targets (#2089)2023-06-28T04:12:49+00:00suleshahid110432064+suleshahid@users.noreply.github.com
This is needed to avoid name clashes and properly build all targets.
The make command names to run the signal tests have changed to use `signal_` before the op.
So:
`test_kernel_fft_test` is now `test_kernel_signal_fft_test`, etc.
BUG=[288938993](<a href="http://b/288938993" rel="nofollow noreferrer noopener" target="_blank">http://b/288938993</a>)https://gitcode.net/xusiwei1236/tflite-micro/-/commit/db846aa96a9bb0c6de39865a8cf5bd308145de75Remove unnecessary defines for Hexagon reference kernel build. (#2091)2023-06-28T16:15:29+00:00Advait Jainadvaitjain@users.noreply.github.com
These were preventing reference kernels from being built with `TARGET=hexagon`. See <a href="http://b/281821666#comment5" rel="nofollow noreferrer noopener" target="_blank">http://b/281821666#comment5</a> for additional context.
Manually tested that the following command passes with this change:
```
make -f tensorflow/lite/micro/tools/make/Makefile TARGET=hexagon -j8 test
```
And without this change we get the following error:
```
tensorflow/lite/micro/kernels/fully_connected.cc:74:49: error: no member named 'filter_buffer_index' in 'tflite::OpDataFullyConnected'
&data->filter_buffer_index);
~~~~ ^
tensorflow/lite/micro/kernels/fully_connected.cc:129:55: error: no member named 'filter_buffer_index' in 'tflite::OpDataFullyConnected'
context->GetScratchBuffer(context, data.filter_buffer_index));
~~~~ ^
```
Note that we do not have any CI coverage for reference kernels with Hexagon since the expectation is that everyone should use the optimized kernels.
However, we still provide a best-effort attempt to explicitly use the reference kernels, if needed.
BUG=<a href="http://b/281821666" rel="nofollow noreferrer noopener" target="_blank">http://b/281821666</a>https://gitcode.net/xusiwei1236/tflite-micro/-/commit/1ec11efcbb362b4e2c7d8ff7bd71c28dff98b610Update docker containers for Xtensa workflows. (#2095)2023-06-29T22:03:45-07:00Advait Jainadvaitjain@users.noreply.github.com
As part of reducing the size of the docker containers for the Xtensa tests, we have split them up based on different versions of the toolchain.
This PR changes the github workflows to use the smaller and more targeted docker containers.
Couple of things to note:
* This PR is not pinning to a specific version of the docker container. This will make it easier to push updates if needed.
* the names of the containers might change again. We want to get somethign going quickly so that CI is green again and PRs can be merged.
* This PR will be merged bypassing the branch protection rules since the Xtensa workflows run as pull_request_target. We will only be able to test the changes in this PR after the merge.
BUG=<a href="http://b/289098887" rel="nofollow noreferrer noopener" target="_blank">http://b/289098887</a>https://gitcode.net/xusiwei1236/tflite-micro/-/commit/4510f61f8abe7c144951cc7a08c954697d32c285Add missing tag to docker image. (#2097)2023-06-29T22:10:32-07:00Advait Jainadvaitjain@users.noreply.github.com
I got this wrong in #2095. Turns out that I need to specifcy the tag.
BUG=<a href="http://b/289098887" rel="nofollow noreferrer noopener" target="_blank">http://b/289098887</a>https://gitcode.net/xusiwei1236/tflite-micro/-/commit/7a3ef35d1459986127e0ff13702e487d66aa6e70Remove leftover riscv32_mcu files (#2090)2023-06-30T08:01:41+00:00RJ Ascanirjascani@google.com
There were a few leftover files after PR #2061 for the riscv32_mcu target. Since that target is no longer buildable, these files are not needed. This PR removes those remaining files.
BUG=<a href="http://b/286622884" rel="nofollow noreferrer noopener" target="_blank">http://b/286622884</a>https://gitcode.net/xusiwei1236/tflite-micro/-/commit/7125afac4d71ff7fe90abb5b198d882ab635a9a0Update the train micro speech model code for Tensorflow 2.x compatibi… (#2098)2023-06-30T08:32:45+00:00pjpratik118897289+pjpratik@users.noreply.github.com
…lity
The latest colab does not support running 1.x version. The code has been updated to be able to run with 2.x version in the colab.
Thanks.
BUG=cleanuphttps://gitcode.net/xusiwei1236/tflite-micro/-/commit/13cd6d82406728fe38880de20933e765ebe0b59dAutomated sync from github.com/tensorflow/tensorflow (#2093)2023-06-30T17:25:43+00:00TFLM-bottflm-github-bot@google.com
BUG=automated sync from upstream
NO_CHECK_TFLITE_FILES=automated sync from upstream