| Age | Commit message (Collapse) | Author |
|---|
|
If a file has the right extension, MLIA previously tried
to load files with invalid content, resulting in confusing
errors. This patch adds better reporting for that scenario
Resolves: MLIA-1051
Signed-off-by: Annie Tallund <annie.tallund@arm.com>
Change-Id: I3f1fd578906a73a58367428f78409866f5da7836
|
|
If the input model for rewriting is quantized:
- Record de-quantized TFRecords
- enable writing de-quantized calibration data for the training
- re-generate augmented training data, if needed
- Use quantization-aware training (QAT) to train the replacement models
- Check if replacement model is quantized:
If source model is quantized, we make sure rewrite's output model
is quantized too. Right now, only int8 is supported so raising
an error if any other datatype is present in the output.
Resolves: MLIA-907, MLIA-908, MLIA-927
Signed-off-by: Benjamin Klimczak <benjamin.klimczak@arm.com>
Change-Id: Icb4070a9e6f1fdb5ce36120d73823986e89ac955
|
|
- Fix input shape of rewrite replacement:
During and after training of the replacement model for a rewrite the
Keras model is converted and saved in TensorFlow Lite format. If the
input shape does not match the teacher model exactly, e.g. if the
batch size is undefined, the TFLiteConverter adds extra operators
during conversion.
- Fix rewritten model output
- Save the model output with the rewritten operator in the output dir
- Log MAE and NRMSE of the rewrite
- Remove 'verbose' flag from rewrite module and rely on the logging
mechanism to control verbose output.
- Re-factor utility classes for rewrites
- Merge the two TFLiteModel classes
- Move functionality to load/save TensorFlow Lite flatbuffers to
nn/tensorflow/tflite_graph
- Fix issue with unknown shape in datasets
After upgrading to TensorFlow 2.12 the unknown shape of the
TFRecordDataset is causing problems when training the replacement models
for rewrites. By explicitly setting the right shape of the tensors we
can work around the issue.
- Adapt default parameters for rewrites. The training steps especially
had to be increased significantly to be effective.
Resolves: MLIA-895, MLIA-907, MLIA-946, MLIA-979
Signed-off-by: Benjamin Klimczak <benjamin.klimczak@arm.com>
Change-Id: I887ad165aed0f2c6e5a0041f64cec5e6c5ab5c5c
|
|
- Update version dependencies in the tox.ini
- Fix linter issues
Change-Id: I04c3a841ee2646a865dab037701d66c28792f2a4
Signed-off-by: Benjamin Klimczak <benjamin.klimczak@arm.com>
|
|
Use "TensorFlow Lite" instead of "TFLite" in
documentation and comments
Change-Id: Ie4450d72fb2e5261d152d72ab8bd94c3da914c46
|
|
Re-factoring the code base to further merge the old AIET code into MLIA.
- Remove last traces of the backend type 'tool'
- Controlled systems removed, including SSH protocol, controller,
RunningCommand, locks etc.
- Build command / build dir and deploy functionality removed from
Applications and Systems
- Moving working_dir()
- Replace module 'output_parser' with new module 'output_consumer' and
merge Base64 parsing into it
- Change the output consumption to optionally remove (i.e. actually
consume) lines
- Use Base64 parsing in GenericInferenceOutputParser, replacing the
regex-based parsing and remove the now unused regex parsing
- Remove AIET reporting
- Pre-install applications by moving them to src/mlia/resources/backends
- Rename aiet-config.json to backend-config.json
- Move tests from tests/mlia/ to tests/
- Adapt unit tests to code changes
- Dependencies removed: paramiko, filelock, psutil
- Fix bug in corstone.py: The wrong resource directory was used which
broke the functionality to download backends.
- Use f-string formatting.
- Use logging instead of print.
Change-Id: I768bc3bb6b2eda57d219ad01be4a8e0a74167d76
|
