History

Haoliang Zhang c58a667812 [Fix] fix the logic when comparing the runtime version strings. PiperOrigin-RevId: 264948915		2019-08-23 17:09:46 -07:00
..
g3doc
graph_transformations	Fix MatMul with transpose_a	2019-08-05 22:21:00 -07:00
python	Make TFLiteConverter build with MLIR internally by default.	2019-07-19 17:22:25 -07:00
runtime
tensorflow_graph_matching	Cleanup usage of the protobuf workspace. See #19032	2019-06-19 10:36:33 -07:00
tflite	[Fix] fix the logic when comparing the runtime version strings.	2019-08-23 17:09:46 -07:00
allocate_transient_arrays.cc
allocate_transient_arrays.h
args.cc
args.h	TFL micro: Add allow_dynamic_tensor flag to TOCO.	2019-07-07 23:37:40 -07:00
BUILD	Create the initial BUILD file for tensorflow/core/platform folder	2019-08-05 14:17:12 -07:00
dump_graphviz.cc	Documentation correction	2019-03-21 16:31:02 +05:30
dump_graphviz.h	Multiple enhancements to graphviz output:	2019-02-11 11:40:41 -08:00
export_tensorflow.cc	Fix SplitV support.	2019-06-13 20:05:16 -07:00
export_tensorflow.h
format_port.h	HAS_GLOBAL_STRING is no longer defined. ::string resolves to std::string on PLATFORM_GOOGLE.	2019-05-30 19:00:50 -07:00
import_tensorflow_test.cc	Give out warning if model has place holder but not in the input_array specified by the user.	2019-06-05 20:00:06 -07:00
import_tensorflow.cc	Merge pull request #31075 from Huawei-MRC-OSI:patch-1	2019-07-29 13:36:19 -07:00
import_tensorflow.h	Automated rollback of commit `4ee64e012a`	2019-04-22 23:03:40 -07:00
model_cmdline_flags_test.cc	TFL: Make the test more robust.	2019-07-02 23:24:00 -07:00
model_cmdline_flags.cc	Updated model_cmdline_flags.cc	2019-02-26 08:51:11 +05:30
model_cmdline_flags.h
model_flags.proto	Explicitly passing the number of dimensions for RNN states.	2019-02-28 13:09:40 -08:00
model.h	Explicitly include necessary headers.	2019-08-01 21:13:35 -07:00
README.md
tensorflow_util.cc
tensorflow_util.h
toco_cmdline_flags_test.cc	TFL micro: Add allow_dynamic_tensor flag to TOCO.	2019-07-07 23:37:40 -07:00
toco_cmdline_flags.cc	TFL micro: Add allow_dynamic_tensor flag to TOCO.	2019-07-07 23:37:40 -07:00
toco_cmdline_flags.h
toco_convert_test.cc	Give out warning if model has place holder but not in the input_array specified by the user.	2019-06-05 20:00:06 -07:00
toco_convert.cc	Fewer TOCO crashes.	2019-02-06 13:13:59 -08:00
toco_convert.h
toco_flags.proto	TFL micro: Add allow_dynamic_tensor flag to TOCO.	2019-07-07 23:37:40 -07:00
toco_graphviz_dump_options.cc
toco_graphviz_dump_options.h
toco_port_test.cc
toco_port.cc
toco_port.h
toco_tooling.cc	TFL micro: Add allow_dynamic_tensor flag to TOCO.	2019-07-07 23:37:40 -07:00
toco_tooling.h	Fewer TOCO crashes.	2019-02-06 13:13:59 -08:00
toco_types.h
toco.cc	TOCO: Print error message if conversion fails.	2019-02-10 11:35:12 -08:00
tooling_util_test.cc
tooling_util.cc	support MatrixSetDiagV2 op	2019-07-02 14:27:47 -07:00
tooling_util.h	Fix the deletion of ops and their arrays in toco:	2019-06-14 13:37:29 -07:00
types.proto	Propagate half precision float through tflite	2019-05-07 15:01:57 -07:00

README.md

TensorFlow Lite Converter

The TensorFlow Lite Converter converts TensorFlow graphs into TensorFlow Lite graphs. There are additional usages that are also detailed in the usage documentation.

Usage documentation

Usage information is given in these documents:

Where the converter fits in the TensorFlow landscape

Once an application developer has a trained TensorFlow model, the TensorFlow Lite Converter will accept that model and generate a TensorFlow Lite FlatBuffer file. The converter currently supports SavedModels, frozen graphs (models generated via freeze_graph.py), and tf.Keras model files. The TensorFlow Lite FlatBuffer file can be shipped to client devices, generally mobile devices, where the TensorFlow Lite interpreter handles them on-device. This flow is represented in the diagram below.