- 16 6月, 2021 1 次提交
-
-
由 Advait Jain 提交于
* Add Hexagon optimized kernels. * Hexagon optimized kernels copied from https://source.codeaurora.org/quic/embedded_ai/tensorflow at 2d052806c211144875c89315a4fc6f1393064cf6 * Changed the include paths and directory structure a bit. * Modified Makefile to allow optimized kernels to be in a separate directory * Path to the Hexagon lib is now specified on the command line. Verified that the optimized kernels are properly built and linked with: ``` make -f tensorflow/lite/micro/tools/make/Makefile TARGET=hexagon OPTIMIZED_KERNEL_DIR=hexagon OPTIMIZED_KERNEL_DIR_PREFIX=third_party HEXAGON_TFLM_LIB=~/Qualcomm/tflm_google/hexagon_tflm_core.a -j8 run_keyword_benchmark ``` Gives: ``` KeywordRunNIerations(1) took 52608 ticks (52 ms) ``` Whereas reference kernels with: ``` make -f tensorflow/lite/micro/tools/make/Makefile TARGET=hexagon -j8 run_keyword_benchmark ``` Gives: ``` KeywordRunNIerations(1) took 110248 ticks (110 ms) ``` BUG=http://b/190754463 * enable code style checks for third_party as well. * clang-formatted the hexagon kernels. * Rename to keep scope focused on what we currently need.
-
- 15 6月, 2021 2 次提交
-
-
由 Advait Jain 提交于
We are only running these checks as part of the GitHub actions CI and are thus always guaranteed to be operating from a git repository.
-
由 TFLM-bot 提交于
* Sync from upstream TF. * Manually pull in TFLM changes from b050465c1de2d2d633655a83e504d8fbf86edb9b These are necessary to have all the required checks pass and will fix #164 Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
- 14 6月, 2021 1 次提交
-
-
由 Jens Elofsson 提交于
* Fix broken links to images in preallocated tensors RFC. * Update links to refer to the new repo. * Add line breaks to make it easier to read. Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
- 11 6月, 2021 1 次提交
-
-
由 Advait Jain 提交于
-
- 10 6月, 2021 5 次提交
-
-
由 Advait Jain 提交于
Following the pattern of #151 and fixes #11.
-
由 Advait Jain 提交于
* Remove -Wno-strict-aliasing * patch stm32_bare_lib to remove the unused function that was the reason why we were disabling -Wno-strict-aliasing for certain targets. * This is general cleanup stemming from https://github.com/tensorflow/tflite-micro/pull/109#issuecomment-856858010 * add -Wno-strict-aliasing for CMSIS.
-
由 Advait Jain 提交于
* make srcs not executable * make bash script executable
-
由 David Davis 提交于
When using the MicroInterpreter, the TfLiteEvalTensor must be the source of truth for tensor dims. Additional fix for issue #152 Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 David Davis 提交于
Also check that the AllocatePersistentBuffer method is available. Additional fix for issue #152 Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
- 09 6月, 2021 2 次提交
-
-
由 Måns Nilsson 提交于
* Revert "Revert: PR #48758: TFLM: Remove uint8 support for CMSIS-NN kernels" This reverts commit 6d8a184710164c7bcc41d3bd29fe4ef46836b6c9. * TFLM: Increase stack size for bluepill Change-Id: Ie278c6fff55cc620104c69167273a92a2d8e639a * Update PPD unit test * Remove leftovers from when PPD supported dynamic output dimensions. * Correct number of output elements. * Correct number of input elements for last unit test. Change-Id: I66fd36d7eb48816986450033bc83c6b240d559b6 * Revert "TFLM: Increase stack size for bluepill" This reverts commit c7fc2a0e0f16d2ab10c44389033822503c178e69. With the fix to the PPD unit test from 6e39d12d7688143bd17acb2c5d422385baef9e0f, increasing the bluepill stack size should not be necessary. * Fix circular buffer formatting. The CI check for clang-format needs to be fixed. Co-authored-by: NAdvait Jain <advaitjain@google.com> Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 mike dooley 提交于
support both 2 second and 4 second input size models
-
- 08 6月, 2021 5 次提交
-
-
由 TFLM-bot 提交于
-
由 yorange1 提交于
-
由 TFLM-bot 提交于
-
由 David Davis 提交于
Added batch_matmul.h and tensor_utils_common.h that are needed for the BATCH_MATMUL operator. Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 Advait Jain 提交于
This is required only because some of the third_party_downloads are still using the Makefile functions and thus require an explicit `make third_party_downloads`. Additionally, we sometimes also want to avoid all downloads and this PR adds in a flag for that as well. Fixes #119 and the corresponding internal bug http://b/184886633 Related bugs are http://b/143904317 and http://b/183229719
-
- 05 6月, 2021 3 次提交
-
-
由 Advait Jain 提交于
-
https://github.com/tensorflow/tensorflow/pull/47935由 Advait Jain 提交于
* Revert "Disable currently failing test to make the Xtensa build green. (#124)" This reverts commit 20cdd3fb. * Cherry pick https://github.com/tensorflow/tensorflow/pull/50072 * clang-formatting fix.
-
由 Advait Jain 提交于
* Trigger workflow from pull_request_target. * Explicitly call third_party_downloads prior to running the tests. BUG=http://b/190108540
-
- 04 6月, 2021 5 次提交
-
-
由 Advait Jain 提交于
-
由 Advait Jain 提交于
BUG=http://b/190108540
-
由 Advait Jain 提交于
* Disable currently failing test to make the Xtensa build green. BUG=https://github.com/tensorflow/tflite-micro/issues/123 * excluded some additional tests that were failing for Hifi5 and Vision P6. * formatting fixes. These are part of an already merged PR. We do not have proper checks for the formatting as part of the github actions and hence these were only caught by my local git hooks.
-
由 jwithers 提交于
Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 bhanuprakashbv 提交于
* Integrate the HiFi 5 Neural Network Library v1.5.0 Integrated HiFi 5 NN Library kernels for below operators svdf(int8), quantize(int8), add(int8, uint8 and float32), mul(int8, uint8 and float32), relu(int8, uint8 and float32) and relu6(int8, uint8 and float32) Signed-off-by: NAnirban Mandal <amandal@cadence.com> Signed-off-by: NBiswa Dhal <bdhal@cadence.com> Signed-off-by: NBhanu Prakash Bandaru Venkata <bhanup@cadence.com> Signed-off-by: NChaitanya Sanjay Muley <cmuley@cadence.com> Signed-off-by: NHarinarayanan E V <hariev@cadence.com> Signed-off-by: NHarshavardhan Ravindra Joshi <joshih@cadence.com> * Integrated HiFi 5 NN Library kernels for below operators svdf(int8), quantize(int8), Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
- 03 6月, 2021 3 次提交
-
-
由 TFLM-bot 提交于
-
由 TFLM-bot 提交于
-
由 Ryan Kuester 提交于
Add the header containing the reference implementation of op SPACE_TO_DEPTH to the list of files which are synced from the upstream TensorFlow repository. This reference is needed for the forthcoming implementation of SPACE_TO_DEPTH here in micro.
-
- 02 6月, 2021 2 次提交
-
-
由 Advait Jain 提交于
* The TFLM Makefile globs all the shared TFL/TFLM code. * This allows us to move the explicit list of sources and headers to the sync script since that is where we determine what code needs to be sync'd from upstream TF. With this change, we are ready to have the tflite_micro repository be the source of truth for all TFLM code and sync only the shared TFL/TFLM code from the tensorflow repo. Bug: http://b/182914089
-
由 TFLM-bot 提交于
-
- 28 5月, 2021 3 次提交
-
-
由 TFLM-bot 提交于
-
由 Advait Jain 提交于
-
由 Advait Jain 提交于
I have moved the Xtensa continuous builds to pull from the tflite-micro repo and this change is adding a corresponding build badge.
-
- 27 5月, 2021 2 次提交
-
-
由 TFLM-bot 提交于
-
由 Advait Jain 提交于
Bug: http://b/184876027
-
- 26 5月, 2021 1 次提交
-
-
由 TFLM-bot 提交于
-
- 25 5月, 2021 4 次提交
-
-
由 TFLM-bot 提交于
-
由 Advait Jain 提交于
* remove unnecessary triggers. * explicitly specify branch for pull request trigger
-
由 rsun-bdti 提交于
* Port TFL kernel Gather to TFL Micro * Remove TODO comment in micro/kernel/gather.cc Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 Advait Jain 提交于
-