| .. |
|
calibration
|
Integer LSTMs: Name scratch arrays Based on what gate they are representing. Make naming consistent with float/hybrid versions.
|
2020-06-19 19:33:41 -07:00 |
|
g3doc
|
|
|
|
python
|
Fix misspelling
|
2020-04-29 21:10:09 +09:00 |
|
sparsity
|
Fix a bug in the format converter.
|
2020-04-29 22:04:35 -07:00 |
|
testdata
|
Merge pull request #36251 from wwwind:interface_16x8
|
2020-06-18 19:47:52 -07:00 |
|
BUILD
|
Merge pull request #36251 from wwwind:interface_16x8
|
2020-06-18 19:47:52 -07:00 |
|
model_utils_test.cc
|
Strengthen IsQuantized rule to actually check quantization params rather than just looking at the type.
|
2019-06-17 09:55:06 -07:00 |
|
model_utils.cc
|
Modify op version in optimize only if convertor version < quantized version.
|
2020-05-12 15:44:28 -07:00 |
|
model_utils.h
|
Support unknown dimensions in quantized models.
|
2020-05-06 18:30:12 -07:00 |
|
modify_model_interface_main.cc
|
fix the typo
|
2020-04-22 17:31:05 -07:00 |
|
modify_model_interface_test.cc
|
Update test case for modify model interface. The input and output float tensors are in the beginning and end of the model respectively.
|
2020-06-08 13:12:55 -07:00 |
|
modify_model_interface.cc
|
Add a safeguard for tensor removal. If a tensor to be removed is at the beginning of the tensor list, keep it.
|
2020-06-02 16:07:26 -07:00 |
|
modify_model_interface.h
|
Create a helper function to change a float models's interface to uint8. This is for users to use on inputs, rather than relying on infererence_input and inference_output type in the 2.0 converter.
|
2020-03-23 09:51:46 -07:00 |
|
operator_property.cc
|
Full int8 quantization BatchMatMul
|
2020-06-19 07:33:08 -07:00 |
|
operator_property.h
|
Merge branch 'master' into interface_16x8
|
2020-06-02 10:44:46 +01:00 |
|
quantization_utils_test.cc
|
Added an option TFLITE_BUILTINS_ACTIVATIONS_INT16_WEIGHTS_INT8 to
|
2020-01-27 14:01:11 +00:00 |
|
quantization_utils.cc
|
Merge pull request #36251 from wwwind:interface_16x8
|
2020-06-18 19:47:52 -07:00 |
|
quantization_utils.h
|
Added an option TFLITE_BUILTINS_ACTIVATIONS_INT16_WEIGHTS_INT8 to
|
2020-01-27 14:01:11 +00:00 |
|
quantization_wrapper_utils_test.cc
|
Refactor test to enable support of other variants of LSTM.
|
2019-11-26 11:38:07 -08:00 |
|
quantization_wrapper_utils.cc
|
Add utility functions for integration. This supports the calibration case that the model is initialized multiple times.
|
2019-11-14 15:03:39 -08:00 |
|
quantization_wrapper_utils.h
|
Add utility functions for integration. This supports the calibration case that the model is initialized multiple times.
|
2019-11-14 15:03:39 -08:00 |
|
quantization_wrapper.cc
|
Merge pull request #36251 from wwwind:interface_16x8
|
2020-06-18 19:47:52 -07:00 |
|
quantization_wrapper.h
|
Add utility functions for integration. This supports the calibration case that the model is initialized multiple times.
|
2019-11-14 15:03:39 -08:00 |
|
quantize_model_test.cc
|
Merge pull request #36251 from wwwind:interface_16x8
|
2020-06-18 19:47:52 -07:00 |
|
quantize_model.cc
|
Merge branch 'upstream/master' into interface_16x8
|
2020-06-08 17:03:06 +01:00 |
|
quantize_model.h
|
Fix for the broken 16-bit interface after latest changes to master.
|
2020-03-27 20:14:56 +00:00 |
|
quantize_weights_test.cc
|
Change comment of the external repo #include file origin.
|
2020-03-23 10:42:53 -07:00 |
|
quantize_weights.cc
|
Support unknown dimensions in quantized models.
|
2020-05-06 18:30:12 -07:00 |
|
quantize_weights.h
|
Support quantization to float16
|
2019-05-17 14:57:42 -07:00 |
|
test_util.cc
|
Merge pull request #36251 from wwwind:interface_16x8
|
2020-06-18 19:47:52 -07:00 |
|
test_util.h
|
Merge pull request #36251 from wwwind:interface_16x8
|
2020-06-18 19:47:52 -07:00 |
