feat(compression): integrate automatic FlatBuffer alignment (#3177)

Enhance compress() function to automatically apply proper
FlatBuffer alignment after compression, eliminating the need for
users to manually run tflite_flatbuffer_align as a separate step.
Use the C++ alignment wrapper internally, as the Python
flatbuffers library doesn't respect force_align schema
attributes.

Keep the API unchanged - compress() still returns a bytearray,
but now the output is properly aligned for the TFLM interpreter.
Update documentation, and build dependencies of the Python
package.

BUG=#3125

Author: rkuester

python/tflite_micro/BUILD

@@ -153,6 +153,7 @@ py_package(
     packages = [
         "python.tflite_micro",
         "tensorflow.lite.micro.compression",
+        "tensorflow.lite.micro.tools",
         "tensorflow.lite.micro.tools.generate_test_for_model",
         "tensorflow.lite.python",
         "tensorflow.lite.tools.flatbuffer_utils",
@@ -162,6 +163,7 @@ py_package(
         ":runtime",
         ":version",
         "//tensorflow/lite/micro/compression",
+        "//tensorflow/lite/micro/tools:tflite_flatbuffer_align",
     ],
 )
 

tensorflow/lite/micro/compression/BUILD

@@ -116,6 +116,7 @@ py_library(
         ":metadata_py",
         ":model_facade",
         ":spec",
+        "//tensorflow/lite/micro/tools:tflite_flatbuffer_align",
         "@absl_py//absl:app",
         "@absl_py//absl/flags",
         "@flatbuffers//:runtime_py",
@@ -140,6 +141,11 @@ py_test(
     srcs = [
         "compress_test.py",
     ],
+    tags = [
+        "noasan",
+        "nomsan",  # Sanitizer symbols don't work with Python extension libs
+        "noubsan",
+    ],
     deps = [
         ":compress",
         ":metadata_py",

tensorflow/lite/micro/compression/compress.py

@@ -19,7 +19,9 @@
 import bitarray
 import bitarray.util
 from dataclasses import dataclass, field
+import os
 import sys
+import tempfile
 from typing import ByteString, Iterable, Optional
 
 import absl.app
@@ -30,6 +32,7 @@
 from tflite_micro.tensorflow.lite.micro.compression import model_facade
 from tflite_micro.tensorflow.lite.micro.compression import spec
 from tflite_micro.tensorflow.lite.micro.compression import metadata_py_generated as schema
+from tflite_micro.tensorflow.lite.micro.tools import tflite_flatbuffer_align_wrapper
 
 USAGE = f"""\
 Usage: compress.py --input <in.tflite> --spec <spec.yaml> [--output <out.tflite>]
@@ -250,6 +253,43 @@ def _pack_lookup_tables(tables: list[np.ndarray], table_len: int) -> bytearray:
   return buffer
 
 
+def _apply_flatbuffer_alignment(model_bytes: bytearray) -> bytearray:
+  """Applies proper FlatBuffer alignment to a model.
+  
+  The Python flatbuffers library doesn't respect `force_align` schema attributes,
+  so we use the C++ wrapper which properly handles alignment requirements.
+  
+  Args:
+    model_bytes: The model flatbuffer to align
+    
+  Returns:
+    The properly aligned model flatbuffer
+  """
+  # C++ wrapper requires file paths, not byte buffers
+  with tempfile.NamedTemporaryFile(suffix='.tflite', delete=False) as temp_in:
+    temp_in.write(model_bytes)
+    temp_in_path = temp_in.name
+
+  with tempfile.NamedTemporaryFile(suffix='.tflite', delete=False) as temp_out:
+    temp_out_path = temp_out.name
+
+  try:
+    # Unpack and repack with proper alignment
+    tflite_flatbuffer_align_wrapper.align_tflite_model(temp_in_path,
+                                                       temp_out_path)
+
+    with open(temp_out_path, 'rb') as f:
+      aligned_model = bytearray(f.read())
+
+    return aligned_model
+  finally:
+    # Clean up temporary files
+    if os.path.exists(temp_in_path):
+      os.unlink(temp_in_path)
+    if os.path.exists(temp_out_path):
+      os.unlink(temp_out_path)
+
+
 def compress(model_in: ByteString, specs: Iterable[spec.Tensor]) -> bytearray:
   """Compresses a model .tflite flatbuffer.
 
@@ -291,7 +331,9 @@ def compress(model_in: ByteString, specs: Iterable[spec.Tensor]) -> bytearray:
   # add compression metadata to model
   model.add_metadata(TFLITE_METADATA_KEY, metadata.compile())
 
-  return model.compile()
+  # Compile the model and apply proper alignment
+  unaligned_model = model.compile()
+  return _apply_flatbuffer_alignment(unaligned_model)
 
 
 def _fail_w_usage() -> int:

tensorflow/lite/micro/docs/compression.md

@@ -323,11 +323,6 @@ Once the `YAML` specification is ready, compress the model using the following:
 bazel run -s tensorflow/lite/micro/compression:compress -- --input=binned.tflite --output=compressed.tflite --spec=spec.yaml
 ```
 
-Then align the model:
-```
-bazel run -s tensorflow/lite/micro/tools:tflite_flatbuffer_align -- compressed.tflite compressed_and_aligned.tflite
-```
-
 # The Generic Benchmark Application
 
 The Generic Benchmark Application can be used to see the size of the model, the