- 13 6月, 2023 2 次提交
-
-
由 Advait Jain 提交于
BUG=cleanup
-
由 TFLM-bot 提交于
BUG=automated sync from upstream NO_CHECK_TFLITE_FILES=automated sync from upstream
-
- 12 6月, 2023 1 次提交
-
-
由 Advait Jain 提交于
Internal checks caught this unused import in the BUILD file. BUG=cleanup
-
- 10 6月, 2023 1 次提交
-
-
由 Advait Jain 提交于
BUG=http://b/286456378
-
- 09 6月, 2023 3 次提交
-
-
由 Ryan Kuester 提交于
Rename the Python package `tflm_runtime` to simply `runtime` in preparation for adding it under the new namespace package `tflite_micro`. Its full name will then be `tflite_micro.runtime`. We have kept the `tflm_runtime` target as an alias in order to stage this change. More details in http://b/286456378 BUG=part of #1484
-
由 Advait Jain 提交于
BUG=cleanup
-
由 Ryan Kuester 提交于
Remove the pybind_library() used only as a dependency to the pybind_extension(), and instead simply make the pybind_extension() directly. A pybind_library() may be useful if used in other palces in addition to a pybind_extension(), but with no such use in our tree, it just adds complication. BUG=part of #1484
-
- 08 6月, 2023 2 次提交
-
-
由 Ryan Kuester 提交于
Clarify the guidance for linking a PR description to its bug number. The last bullet was obsolete and confusing now that we require the specific text `BUG=#nn`, as was mentioned in the second bullet. BUG=see description
-
由 TFLM-bot 提交于
BUG=automated sync from upstream NO_CHECK_TFLITE_FILES=automated sync from upstream
-
- 07 6月, 2023 1 次提交
-
-
由 TFLM-bot 提交于
BUG=automated sync from upstream NO_CHECK_TFLITE_FILES=automated sync from upstream
-
- 02 6月, 2023 2 次提交
-
-
由 Advait Jain 提交于
Changes in strided slice from http://cl/536336953 required updates to the kernel test. Note that the TFLM strided\_slice kernel does not support the new offset parameter from http://cl/536336953. See http://b/285340415 for additional context. NO_CHECK_TFLITE_FILES=manual sync from upstream. BUG=http://b/285340415
-
由 David Davis 提交于
@tensorflow/micro Eval method now uses TfLiteEvalTensor. The `has_low_rank_input_condition` variable has been removed from OpData as SELECT_V2 does not use this variable. All code outside of the registration has been placed into the anonymous namespace. bug=fixes #2014
-
- 01 6月, 2023 1 次提交
-
-
由 David Davis 提交于
@tensorflow/micro Fix logic error in Prepare that prevented deallocation of temporary tensors, when scalars were present in the inputs. Added 3 tests from TfLite for scalar inputs. bug=fixes #2010
-
- 31 5月, 2023 1 次提交
-
-
由 RJ Ascani 提交于
The Dequantize op only supports an output type of Float32. This change removes output_multiplier adjustments for Int32 outputs, as it cannot be reached due to an earlier check. It also eliminates logging for unsupported output types in DequantizeEval, as those would also be caught by an earlier check. BUG=http://b/230890286
-
- 29 5月, 2023 1 次提交
-
-
由 David Davis 提交于
@tensorflow/micro Port the EMBEDDING_LOOKUP kernel operator to TFLM. Recreate tests for kernel operator (INT8, FLOAT32). Add SymmetricScaleFromMinMax method to test_helpers.h file. Update Makefiles. Update Bazel build script. bug=fixes #1871
-
- 28 5月, 2023 2 次提交
-
-
由 TANMAY DAS 提交于
BUG=http://b/284589445
-
-
- 27 5月, 2023 3 次提交
-
-
由 TANMAY DAS 提交于
BUG=http://b/280093506
-
由 Måns Nilsson 提交于
BUG=Reference to all ops resolver should be removed
-
由 Steven Toribio 提交于
add 6 test ported from TfLite that test fused activation for Strided Slice [google3 cl](https://critique.corp.google.com/cl/534210019) BUG=[b/254658480](https://b.corp.google.com/254658480)
-
- 26 5月, 2023 2 次提交
-
-
由 TANMAY DAS 提交于
TFLM swapping TfLiteRegistration_V1 struct to a new more embedded friendly api in the TFLMRegistration struct (#2001) BUG=http://b/279811848
-
由 suleshahid 提交于
We need additional visibility into these files for some other tests. BUG=[274635545](http://b/274635545)
-
- 25 5月, 2023 1 次提交
-
-
由 RJ Ascani 提交于
The bazel install script path was changed from ci/install_bazel.sh to ci/install_bazelisk.sh. This commit updates the CONTRIBUTING doc with the new path. BUG=doc fix
-
- 24 5月, 2023 3 次提交
-
-
由 alankelly 提交于
This is so that a new field may safely be added. BUG=http://b/260574895
-
由 Måns Nilsson 提交于
BUG=hello world example has changed name from test_hello_world_test to test_evaluate_cc_test
-
由 Artem Dementyev 提交于
adding simple ops to requantize_flatbuffer. BUG=http://b/283242131
-
- 23 5月, 2023 1 次提交
-
-
由 TANMAY DAS 提交于
BUG=http://b/282976218 Command used to generate the test and verifyed it is working : ``` bazel-bin/tensorflow/lite/micro/integration_tests/generate_per_layer_tests --input_tflite_file=/tmp/seanet_cca_quantized.tflite --output_dir=tensorflow/lite/micro/integration_tests/seanet/transpose_conv ```
-
- 20 5月, 2023 2 次提交
-
-
由 Steven Toribio 提交于
when debugging integration_test generation accidentally changed file permisions of all the integration_test files to xtext(an excetuable text files) when debugging an issue. This pr reverts all the file permissions back to what they should be. Command used ``` chmod --reference=google3_path/integration_test/seanet/kernel/file tflite_micro/integration_test/seanet/kernel/*``` chmod --reference copys file permission from the listed reference file to the 2nd one provided in the arg is currently blocking submit of g3 [sync cl](https://critique.corp.google.com/cl/522420814/analysis) BUG=[b/283542361](https://b.corp.google.com/issues/283542361)
-
由 TFLM-bot 提交于
BUG=automated sync from upstream NO_CHECK_TFLITE_FILES=automated sync from upstream
-
- 19 5月, 2023 3 次提交
-
-
由 Steven Toribio 提交于
Adds TFLMRegistration to newly created MicroCommon.h First change in set of changes to move TFLM entirely away from TfLite Registration struct and add reset into TFLM explained @ the bottom of the following [Doc](https://docs.google.com/document/d/1giWrZQk1DnC1v-9uDDoos2zNBO06SLI7RatsOMu3RY8/edit?resourcekey=0-wkLBasqHyTZm3POas69tLQ#) BUG=[b/279811848](https://b.corp.google.com/issues/279811848)
-
由 suleshahid 提交于
We have an unused pybind rule in the micro/tools BUILD file. BUG=cleanup
-
由 TFLM-bot 提交于
BUG=automated sync from upstream NO_CHECK_TFLITE_FILES=automated sync from upstream
-
- 17 5月, 2023 4 次提交
-
-
由 Steven Toribio 提交于
changes python op resolver visibility to allow for nightly build generated integration test to be able to use it as well. Currently only seanet integration_tests have visibility [test via running workflow on forked repo](https://github.com/turbotoribio/tflite-micro/actions/runs/4997829269) BUG=[b/276918689 ](https://b.corp.google.com/issues/276918689#comment17)
-
由 suleshahid 提交于
The `tflite_flatbuffer_align` tool is currently a C++ script. The functionality of this script is only useful after you have modified the tflite flatbuffer via the Python API, then you would call this script to realign via the C++ API. It would make more sense to wrap this tool in python, and use it directly in the python script, instead of having a two step procedure. We leave a py binary with the same bazel target +args structure to keep support for any prior usecases. BUG=[274637329](http://b/274637329)
-
由 TFLM-bot 提交于
BUG=automated sync from upstream NO_CHECK_TFLITE_FILES=automated sync from upstream
-
由 Advait Jain 提交于
PR #1641 moved all the issue generation into a single workflow. However, that changed missed removing the code in generate_integration_tests.yml. Also, removed some trailing whitespace in the workflow files. BUG=fix issue creation on nightly workflow failure.
-
- 16 5月, 2023 1 次提交
-
-
由 Steven Toribio 提交于
This PR does the following: This cl finishes the depreciation of AllOpsResolver in TFLM and then regenerate generate_per_layer test using PythonOpsResolver instead of AllOpsResolver after adding it to the MakoTemplate. [command used to regenerate integration_test](https://b.corp.google.com/issues/245770006#comment1) [doc for changes](https://docs.google.com/document/d/1vafCGaFGGBauAssAdFABtmgpiRju2V3isioZfLMgeEc/edit?resourcekey=0-pc-aJl3TPZ7NcO9RxrNsrA#) BUG=[b/276918689](https://b.corp.google.com/issues/276918689)
-
- 14 5月, 2023 1 次提交
-
-
由 Ryan Kuester 提交于
Fix the incorrect function comment on KernelRunner::Invoke(), which says it calls a kernel registration's init and prepare---implying the call InitAndPrepare is optional. In fact, Invoke() does not call init and prepare; it only calls invoke. Bug=see description
-
- 13 5月, 2023 2 次提交
-
-
由 Ryan Kuester 提交于
Add an .editorconfig file to set the project's code formatting styles in editors which support the [EditorConfig](https://editorconfig.org) standard. EditorConfig is natively supported by VisualStudio, GitHub, Neovim, etc. BUG=see description
-
由 Steven Toribio 提交于
This PR does the following: 1. Introduces PythonOpsResolver Target 2. removes a reference to AllOpsResolver in remaing places in TFLM repo outside of integration_tests [doc for changes](https://docs.google.com/document/d/1vafCGaFGGBauAssAdFABtmgpiRju2V3isioZfLMgeEc/edit?resourcekey=0-pc-aJl3TPZ7NcO9RxrNsrA#) [google3 cl](https://critique.corp.google.com/cl/531013669) BUG=[b/276918689](https://b.corp.google.com/issues/276918689)
-