Aztec02 commented 9 months ago

Trying to run the hello_world example with bazel ("bazel build tensorflow/lite/micro/examples/hello_world:evaluate") I get the following error:

ERROR: C:/users/userid/appdata/roaming/spb_data/_bazel_userid/zm73hpx2/external/tflm_pip_deps_tensorflow_cpu/BUILD.bazel:22:11: no such package '@tflm_pip_deps_tensorflow_intel//': The repository '@tflm_pip_deps_tensorflow_intel' could not be resolved: Repository '@tflm_pip_deps_tensorflow_intel' is not defined and referenced by '@tflm_pip_deps_tensorflow_cpu//:pkg'

Environment: Windows 10 Bazel 6.4.0 Python 3.10.10 Windows cmd prompt

rascani commented 9 months ago

I wasn't able to reproduce this. Have you modified the python_requirements.txt file? Perhaps just a transient pypi download issue?

Aztec02 commented 9 months ago

I just tried that on another Windows 10 machine with Bazel 6.4.0 and got the exact same error. I have not modified the requirements. I downloaded the bazel-6.4.0-windows-x86_64.exe, renamed it to bazel and added to PATH. Installed the https://aka.ms/vs/17/release/vc_redist.x64.exe Microsoft Visual C++ Redistributable. Cloned the tflite-micro repo. Open the cmd prompt. Typed bazel build tensorflow/lite/micro/examples/hello_world:evaluate and got the exact same error.

I also tried installing the python requirements and running the evaluate.py script manually but there are other package problems there. For example the script imports matplotlib but matplotlib is not in the requirements, etc

Aztec02 commented 9 months ago

Solved this issue by modifying the requirements: In python_requirements.txt changed tensorflow-io-gcs-filesystem==0.32.0 to version 0.31.0 because there is no wheel for windows and version 32 at pypi. Also added tensorflow-intel==2.12.0 \ --hash=sha256:816a6b9018d1ae0defe94508aab2512228fb6a54ab51f5f4f025c857cfe5c7e5

via tensorflow-cpu

as this is a dependency for tensorflow-cpu. I build the hello_world:train target successfully. But then when I try to build hello_world:evaluate target I get a new error "ERROR: An error occurred during the fetch of repository 'local_config_python':". Correct me if I am wrong but this repo needs many modifications to work on windows.

CannoChen commented 9 months ago

I run into a similar problem as you, and there is no good solution yet. I have tried to use Linux System to solve this problem, but I failed...I suspect the official documentation is out of date！！！！ If you have any problemes, Please update in here! I'm sure this will help a lot of people with similar problems.

CannoChen commented 9 months ago

I sloved this problem. Let me tell you how to slove it!!! First, you must create a new virtual env by using conda: conda create -n myenv python=3.10.0 It is import that the version of python must be 3.10!!!!! If your python version is not 3.10, then it will definitely fail!！

secondly, you must install bazel 7.0.

Finally, don't install any packages, just run the command: bazel build tensorflow/lite/micro/examples/micro_speech:audio_preprocessor (you can instead this command for yours~)

Now, everything is ok!!!

Aztec02 commented 9 months ago

Tried in Ubuntu jellyfish and everything builds fine.

liuchangzong commented 9 months ago

Solved this issue by modifying the requirements: In python_requirements.txt changed tensorflow-io-gcs-filesystem==0.32.0 to version 0.31.0 because there is no wheel for windows and version 32 at pypi. Also added tensorflow-intel==2.12.0 --hash=sha256:816a6b9018d1ae0defe94508aab2512228fb6a54ab51f5f4f025c857cfe5c7e5 # via tensorflow-cpu as this is a dependency for tensorflow-cpu. I build the hello_world:train target successfully. But then when I try to build hello_world:evaluate target I get a new error "ERROR: An error occurred during the fetch of repository 'local_config_python':". Correct me if I am wrong but this repo needs many modifications to work on windows.

That's really work!

liuchangzong commented 9 months ago

Solved this issue by modifying the requirements: In python_requirements.txt changed tensorflow-io-gcs-filesystem==0.32.0 to version 0.31.0 because there is no wheel for windows and version 32 at pypi. Also added tensorflow-intel==2.12.0 --hash=sha256:816a6b9018d1ae0defe94508aab2512228fb6a54ab51f5f4f025c857cfe5c7e5 # via tensorflow-cpu as this is a dependency for tensorflow-cpu. I build the hello_world:train target successfully. But then when I try to build hello_world:evaluate target I get a new error "ERROR: An error occurred during the fetch of repository 'local_config_python':". Correct me if I am wrong but this repo needs many modifications to work on windows.

That's really work!

Well, unfortunately, I got another issue when I was trying to "bazel-bin" the .exe file. Something like this:

E:\tinyML\tflite-micro>bazel-bin\tensorflow\lite\micro\examples\hello_world\train --save_tf_model--save_dir=\tmp\model_created`

Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "E:\tinyML\tflite-micro\bazel-bin\tensorflow\lite\micro\examples\hello_world\train.zip__main__.py", line 559, in

File "E:\tinyML\tflite-micro\bazel-bin\tensorflow\lite\micro\examples\hello_world\train.zip__main__.py", line 454, in Main File "E:\tinyML\tflite-micro\bazel-bin\tensorflow\lite\micro\examples\hello_world\train.zip__main__.py", line 204, in CreateModuleSpace

File "E:\tinyML\tflite-micro\bazel-bin\tensorflow\lite\micro\examples\hello_world\train.zip__main__.py", line 190, in ExtractZip

File "D:\Python\Python311\Lib\zipfile.py", line 1660, in extract return self._extract_member(member, path, pwd) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\Python\Python311\Lib\zipfile.py", line 1731, in _extract_member open(targetpath, "wb") as target: ^^^^^^^^^^^^^^^^^^^^^^

FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\MyName\AppData\Local\Temp\Bazel.runfiles_93cfifk4\runfiles\tflm_pip_deps_tensorflow_intel\site-packages\tensorflow\include\external\llvm-project\mlir\_virtual_includes\AffineMemoryOpInterfacesIncGen\mlir\Dialect\Affine\IR\AffineMemoryOpInterfaces.cpp.inc'`

github-actions[bot] commented 8 months ago

"This issue is being marked as stale due to inactivity. Remove label or comment to prevent closure in 5 days."

CannoChen commented 8 months ago