- 10 7月, 2021 2 次提交
-
-
由 yair_ehrenwald 提交于
* fixed scratch buffer extern decleration in ceva_common.h * Add optimized depthwise conv BX1 and SP500 * fixed year in copyright notice Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 Jens Elofsson 提交于
-
- 09 7月, 2021 5 次提交
-
-
由 Advait Jain 提交于
-
由 Shlomi Regev 提交于
- Save ~4KB of RAM by implementing a lightweight version of the vector, which only accesses integer values, without implicit conversion. - Change Micro kernels to access flexbuffers as vectors instead of maps, which is recommended by the flexbuffers doc for efficiency. Since the values in the vector are ordered alphabetically by their keys, the kernels can access them by index instead. - Revert detection_postprocess to the native flexbuffers API. The LiteVector API doesn't support IsNull() and I prefer not to support two custom flexbuffer APIs. Co-authored-by: NNat Jeffries <natmjeffries@gmail.com>
-
由 Advait Jain 提交于
BUG=http://b/192614484
-
由 Advait Jain 提交于
-
由 Jens Elofsson 提交于
* Remove MICROLITE_CC_KERNELS_SRCS from the MICROLITE_CC_SRCS list. Stop the kernel sources from being compiled into both the core objects and kernel objects. * Add kernel sources to list_library_sources * Fix project generation presubmit Co-authored-by: NNat Jeffries <natmjeffries@gmail.com>
-
- 08 7月, 2021 2 次提交
-
-
由 Måns Nilsson 提交于
Change-Id: I23e7425445fd8fdb37be0ebb43cb9c8529ceb368 Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 yair_ehrenwald 提交于
Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
- 07 7月, 2021 1 次提交
-
-
由 TFLM-bot 提交于
-
- 02 7月, 2021 1 次提交
-
-
由 Nat Jeffries 提交于
Previously, some sources were built before THIRD_PARTY_TARGETS was downloaded, meaning dependencies in those sources on flatbuffers, gemmlowp, etc could be missed. An alternative to this "dependency for every source" approach could be to add $(THIRD_PARTY_TARGETS) as the first dependency in $(MICROLITE_LIB_PATH), but this relies on Make dependency ordering which is ill-advised Background: https://stackoverflow.com/questions/9159960/order-of-processing-components-in-makefile
-
- 30 6月, 2021 2 次提交
-
-
由 Nat Jeffries 提交于
Lots of large models have >50 ops, so upping the max number to 1024 seems reasonable. Most TFLM models should not exceed 1024 ops.
-
由 Nat Jeffries 提交于
-
- 29 6月, 2021 1 次提交
-
-
由 TFLM-bot 提交于
-
- 26 6月, 2021 4 次提交
-
-
由 Nat Jeffries 提交于
Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 Nat Jeffries 提交于
-
由 Advait Jain 提交于
Verified that the following commands work: ``` 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 test_kernel_fully_connected_test 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 test_kernel_svdf_test 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 run_keyword_benchmark ``` BUG=http://b/174781826
-
由 Advait Jain 提交于
The TFLM Makefile itself ensures that OPTIMIZED_KERNEL_DIR is a valid path. However, for some special cases (such as http://cl/379964344), we would like to allow the specialization to gracefully handle a directory that does not exist in the tree. BUG=see use-case from http://cl/379964344
-
- 25 6月, 2021 2 次提交
-
-
由 TFLM-bot 提交于
-
由 Shlomi Regev 提交于
- Add external lib references to Hexagon FFT functions
-
- 24 6月, 2021 2 次提交
-
-
由 TFLM-bot 提交于
-
由 Advait Jain 提交于
-
- 23 6月, 2021 4 次提交
-
-
由 Nat Jeffries 提交于
Add support for 8x16 kernels in TFLM. Support 16-bit activations and 8-bit weights for a subset of the TFLM kernels. BUG=http://b/184839019
-
由 Advait Jain 提交于
* Fix the Vision P6 build. Tested the following command: ``` make -f tensorflow/lite/micro/tools/make/Makefile TARGET=xtensa OPTIMIZED_KERNEL_DIR=xtensa TARGET_ARCH=vision_p6 XTENSA_CORE=P6_200528 build -j8 ``` * Fix the Fusion F1 build.
-
由 Advait Jain 提交于
-
由 jwithers 提交于
* check PR description for a BUG= line * Changed file name and added some more documentation. * move the if to be on the steo instead of at the job. This is to allow for the job to succeed (instead of being skipped) when the BUG= text is found. And as a result we can have it be a required status check. Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com> Co-authored-by: NAdvait Jain <advaitjain@google.com>
-
- 22 6月, 2021 3 次提交
-
-
由 Advait Jain 提交于
* share code with the cmsis implementation * groundwork needed to add in Xtensa implementation BUG=https://github.com/tensorflow/tflite-micro/issues/205
-
由 Ryan Kuester 提交于
* Remove lite-specific code from copy of SPACE_TO_DEPTH Remove the bulk of lite-specific code from the micro implementation of operator SPACE_TO_DEPTH. - Flatten namespace - Don't resize output tensors - Remove type other than int8 and float32 - Don't use gtest * Port operator SPACE_TO_DEPTH from lite Port the SPACE_TO_DEPTH operator from lite to micro. Add the operator and test to the build. Co-authored-by: NPete Warden <pete@petewarden.com>
-
由 Patrik Laurell 提交于
* Update ethosu driver entrypoint * Update copyright notice Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
- 19 6月, 2021 1 次提交
-
-
由 Måns Nilsson 提交于
Overrides core optimization level from -Os to -O2.
-
- 18 6月, 2021 7 次提交
-
-
由 Advait Jain 提交于
* Remove hifimini support from the bazel build. * Consolidate all kernel BUILD targets into micro_ops * remove select statements for xtensa_hifimini BUG=http://b/191406930 * fix bazel build.
-
由 Advait Jain 提交于
* Explicitly build all targets with bazel as part of CI. With #194 we removed the build_test targets. We are instead going to explicitly have a bazel build command as part of the CI checks. * fix typo
-
由 Fredrik Knutsson 提交于
* Put location variables before target .inc's are included They can be useful in the target .inc Change-Id: I0ee3f77f79be272f4dc3502fb4a38017d8162fa1 * Add ARMC6 compiler support for Corstone-300 target * Use fromelf instead of objcopy for armclang toolchain when generating a binary * Add ARMC6 linker and build flags to the Corstone-330 target * Add RETARGET macro in patch script to avoid undefined symbol build error for ARMC6. * Use a reduced set of Cortex-M CPU's for easier maintenance. Change-Id: Id9a20d57fa4fa0f1339f44523417e2dabfe7e152 * Review comment - Override exit symbol only for GCC Change-Id: I93628c92ee352f36c7e7dd99351e4a73c29a8d30 * Review comment - Change how to pass linker options for ARMC6 Change-Id: Iacec5a6df6902bc8a14f460d63ae917039f982cc * Review comment - correct the upmerge Change-Id: I9a250f6a444336f9b95428a126a94a18f40060dc Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 Advait Jain 提交于
Since the CI builds all the targets (binary and test), and the source of truth for all the TFLM code is having a build_test is not adding any additional value. Additionally, a build_test seems to add additional downloads which would make it slower to run the bazel CI from within a docker container.
-
由 jwithers 提交于
Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-
由 Advait Jain 提交于
-
由 Advait Jain 提交于
* TFLM refactor out of github.com/tensorflow is mostly complete. * The repo-level readme in the current repository is the starting point for anyone looking at this repo.
-
- 17 6月, 2021 3 次提交
-
-
由 Advait Jain 提交于
This is the OSS counterpart for http://cl/379367435 BUG=http://b/191077991
-
由 Advait Jain 提交于
Similar to https://github.com/actions/checkout/issues/518
-
由 Nat Jeffries 提交于
Co-authored-by: NAdvait Jain <advaitjain@users.noreply.github.com>
-