 View on TensorFlow.org
|
 Run in Google Colab
|
 View source on GitHub
|
 Download notebook
|
TensorFlow Lite Model Analyzer API helps you analyze models in TensorFlow Lite format by listing a model's structure.
Model Analyzer API
The following API is available for the TensorFlow Lite Model Analyzer.
tf.lite.experimental.Analyzer.analyze(model_path=None,
model_content=None,
gpu_compatibility=False)
You can find the API details from https://www.tensorflow.org/api_docs/python/tf/lite/experimental/Analyzer or run help(tf.lite.experimental.Analyzer.analyze) from a Python terminal.
Basic usage with simple Keras model
The following code shows basic usage of Model Analyzer. It shows contents of the converted Keras model in TFLite model content, formatted as a flatbuffer object.
import tensorflow as tf
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(128, 128)),
tf.keras.layers.Dense(256, activation='relu'),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(10)
])
fb_model = tf.lite.TFLiteConverter.from_keras_model(model).convert()
tf.lite.experimental.Analyzer.analyze(model_content=fb_model)
2024-07-19 11:27:29.208422: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-07-19 11:27:29.229966: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-07-19 11:27:29.236377: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
/tmpfs/src/tf_docs_env/lib/python3.9/site-packages/keras/src/layers/reshaping/flatten.py:37: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
super().__init__(**kwargs)
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
I0000 00:00:1721388451.839247 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.842671 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.846232 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.849870 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.861636 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.864671 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.868106 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.871497 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.875019 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.878047 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.881448 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388451.884898 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.110709 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.112668 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.115137 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.117199 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.119200 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.121017 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.122905 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.124906 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.126819 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.128633 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.130518 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.132591 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.169725 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.171619 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.173640 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.175665 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.177603 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.179458 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.181363 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.183332 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.185262 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.187577 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.189950 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388453.192326 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmpyjg71hva/assets
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmpyjg71hva/assets
Saved artifact at '/tmpfs/tmp/tmpyjg71hva'. The following endpoints are available:
* Endpoint 'serve'
args_0 (POSITIONAL_ONLY): TensorSpec(shape=(None, 128, 128), dtype=tf.float32, name='keras_tensor')
Output Type:
TensorSpec(shape=(None, 10), dtype=tf.float32, name=None)
Captures:
140404369422112: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404369422288: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404360324896: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404360324720: TensorSpec(shape=(), dtype=tf.resource, name=None)
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
W0000 00:00:1721388453.946496 13762 tf_tfl_flatbuffer_helpers.cc:392] Ignored output_format.
W0000 00:00:1721388453.946540 13762 tf_tfl_flatbuffer_helpers.cc:395] Ignored drop_control_dependency.
=== TFLite ModelAnalyzer ===
Your TFLite model has '1' subgraph(s). In the subgraph description below,
T# represents the Tensor numbers. For example, in Subgraph#0, the RESHAPE op takes
tensor #0 and tensor #3 as input and produces tensor #4 as output.
Subgraph#0 main(T#0) -> [T#6]
Op#0 RESHAPE(T#0, T#3[-1, 16384]) -> [T#4]
Op#1 FULLY_CONNECTED(T#4, T#2, T#-1) -> [T#5]
Op#2 FULLY_CONNECTED(T#5, T#1, T#-1) -> [T#6]
Tensors of Subgraph#0
T#0(serving_default_keras_tensor:0) shape_signature:[-1, 128, 128], type:FLOAT32
T#1(arith.constant) shape:[10, 256], type:FLOAT32 RO 10240 bytes, buffer: 2, data:[0.0467508, -0.0178563, -0.0864473, 0.0911474, 0.0558964, ...]
T#2(arith.constant1) shape:[256, 16384], type:FLOAT32 RO 16777216 bytes, buffer: 3, data:[0.0183942, -0.0170649, -0.0162605, 0.0170177, -0.002608, ...]
T#3(arith.constant2) shape:[2], type:INT32 RO 8 bytes, buffer: 4, data:[-1, 16384]
T#4(sequential_1/flatten_1/Reshape) shape_signature:[-1, 16384], type:FLOAT32
T#5(sequential_1/dense_1/MatMul;sequential_1/dense_1/Relu;sequential_1/dense_1/Add) shape_signature:[-1, 256], type:FLOAT32
T#6(StatefulPartitionedCall_1:0) shape_signature:[-1, 10], type:FLOAT32
---------------------------------------------------------------
Your TFLite model has '1' signature_def(s).
Signature#0 key: 'serving_default'
- Subgraph: Subgraph#0
- Inputs:
'keras_tensor' : T#0
- Outputs:
'output_0' : T#6
---------------------------------------------------------------
Model size: 16789032 bytes
Non-data buffer size: 1456 bytes (00.01 %)
Total data buffer size: 16787576 bytes (99.99 %)
(Zero value buffers): 0 bytes (00.00 %)
* Buffers of TFLite model are mostly used for constant tensors.
And zero value buffers are buffers filled with zeros.
Non-data buffers area are used to store operators, subgraphs and etc.
You can find more details from https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/schema/schema.fbs
Basic usage with MobileNetV3Large Keras model
This API works with large models such as MobileNetV3Large. Since the output is large, you might want to browse it with your favorite text editor.
model = tf.keras.applications.MobileNetV3Large()
fb_model = tf.lite.TFLiteConverter.from_keras_model(model).convert()
tf.lite.experimental.Analyzer.analyze(model_content=fb_model)
/tmpfs/src/tf_docs_env/lib/python3.9/site-packages/keras/src/applications/mobilenet_v3.py:517: UserWarning: `input_shape` is undefined or non-square, or `rows` is not 224. Weights for input shape (224, 224) will be loaded as the default.
return MobileNetV3(
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v3/weights_mobilenet_v3_large_224_1.0_float.h5
22661472/22661472 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmpfncac30q/assets
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmpfncac30q/assets
Saved artifact at '/tmpfs/tmp/tmpfncac30q'. The following endpoints are available:
* Endpoint 'serve'
args_0 (POSITIONAL_ONLY): TensorSpec(shape=(None, None, None, 3), dtype=tf.float32, name='keras_tensor_5')
Output Type:
TensorSpec(shape=(None, 1000), dtype=tf.float32, name=None)
Captures:
140404092101728: TensorSpec(shape=(), dtype=tf.resource, name=None)
140406936371440: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092739824: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092661600: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092661424: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404091826944: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092569904: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092568320: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092824032: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092821744: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092175984: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092221216: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092222976: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092176864: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092174928: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092199680: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092243808: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092241696: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092201968: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404092243280: TensorSpec(shape=(), dtype=tf.resource, name=None)
140404091841872: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155097904: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155098960: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155096672: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155097728: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155115168: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155126224: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155127280: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155124992: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403155126048: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154618320: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154637392: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154638448: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154619728: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154637216: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154650560: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154669808: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154670864: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154652144: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154669632: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154703456: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154719488: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154719664: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154705040: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154706096: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154736048: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154760272: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154760448: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154737632: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154738688: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154789824: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154805328: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154806384: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154791760: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154805152: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154827568: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154827392: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154851968: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154852144: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154361680: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154385200: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154386256: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154383968: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154385024: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154405680: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154428848: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154429904: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154407088: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154428672: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154454304: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154477648: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154478704: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154455888: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154477472: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154495792: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154495968: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154512176: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154512352: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154550272: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154574320: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154587712: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154573088: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154574144: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154591584: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154611008: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154112752: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154609776: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154610832: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154116272: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154135520: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154136576: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154134288: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154135344: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154178288: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154178112: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154181984: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154206784: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154232592: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154256640: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154257696: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154234880: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154256464: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154282272: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154309712: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154310768: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154283856: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154309536: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154327856: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154339440: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154340496: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154338208: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154339264: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153844704: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153864128: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153865184: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153862896: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153863952: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153884608: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153907776: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153908832: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153886016: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153887072: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153937328: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153961728: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153961904: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153939088: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153940144: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153974016: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153994320: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153994496: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153975776: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153976832: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154023168: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154035104: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154035280: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154024752: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154025808: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154063776: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154079984: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154080160: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154065536: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403154066592: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153592560: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153608768: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153608944: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153594320: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153595376: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153633520: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153636336: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153641536: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153635104: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153636160: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153645408: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153689408: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153699056: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153688176: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153689232: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153702576: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153722000: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153743936: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153720768: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153721824: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153747632: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153766880: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153767936: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153765648: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153766704: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153796432: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153815856: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153816912: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153814624: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153815680: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153309536: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153330240: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153359088: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153358912: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153388992: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153408944: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153410000: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153391280: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153408768: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153421232: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153449376: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153449552: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153422640: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153423696: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153473952: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153490160: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153490336: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153475712: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153476768: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153519712: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153519536: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153540368: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153540192: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153045984: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153070032: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153071088: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153068800: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153069856: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153091568: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153114912: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153115968: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153113680: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153114736: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153141248: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153161024: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153182960: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153159792: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153160848: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153207888: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153207712: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153232640: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153232464: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153262544: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153290688: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153291744: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153289456: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153290512: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152782784: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152793664: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152794720: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152784192: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152785248: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152823216: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152843520: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152843696: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152824976: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152826032: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152864880: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152864704: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152889632: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152889456: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152919536: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152947680: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152948736: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152946448: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152947504: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152965120: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152992560: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152993616: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152991328: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152992384: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153018016: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153033344: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153034400: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153032112: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403153033168: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152535392: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152557856: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152559968: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152584768: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152614672: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152634624: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152635680: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152616960: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152634448: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152647968: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152671312: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152672368: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152649552: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152671136: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152697648: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152697824: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152733280: TensorSpec(shape=(), dtype=tf.resource, name=None)
140403152733104: TensorSpec(shape=(), dtype=tf.resource, name=None)
W0000 00:00:1721388464.327364 13762 tf_tfl_flatbuffer_helpers.cc:392] Ignored output_format.
W0000 00:00:1721388464.327393 13762 tf_tfl_flatbuffer_helpers.cc:395] Ignored drop_control_dependency.
=== TFLite ModelAnalyzer ===
Your TFLite model has '1' subgraph(s). In the subgraph description below,
T# represents the Tensor numbers. For example, in Subgraph#0, the MUL op takes
tensor #0 and tensor #134 as input and produces tensor #136 as output.
Subgraph#0 main(T#0) -> [T#263]
Op#0 MUL(T#0, T#134) -> [T#136]
Op#1 ADD(T#136, T#135) -> [T#137]
Op#2 CONV_2D(T#137, T#128, T#79) -> [T#138]
Op#3 HARD_SWISH(T#138) -> [T#139]
Op#4 DEPTHWISE_CONV_2D(T#139, T#78, T#36) -> [T#140]
Op#5 CONV_2D(T#140, T#127, T#77) -> [T#141]
Op#6 ADD(T#139, T#141) -> [T#142]
Op#7 CONV_2D(T#142, T#126, T#35) -> [T#143]
Op#8 PAD(T#143, T#132[0, 0, 0, 1, 0, ...]) -> [T#144]
Op#9 DEPTHWISE_CONV_2D(T#144, T#76, T#34) -> [T#145]
Op#10 CONV_2D(T#145, T#125, T#75) -> [T#146]
Op#11 CONV_2D(T#146, T#124, T#33) -> [T#147]
Op#12 DEPTHWISE_CONV_2D(T#147, T#74, T#32) -> [T#148]
Op#13 CONV_2D(T#148, T#123, T#73) -> [T#149]
Op#14 ADD(T#146, T#149) -> [T#150]
Op#15 CONV_2D(T#150, T#122, T#31) -> [T#151]
Op#16 PAD(T#151, T#131[0, 0, 1, 2, 1, ...]) -> [T#152]
Op#17 DEPTHWISE_CONV_2D(T#152, T#72, T#30) -> [T#153]
Op#18 MEAN(T#153, T#130[1, 2]) -> [T#154]
Op#19 CONV_2D(T#154, T#121, T#29) -> [T#155]
Op#20 CONV_2D(T#155, T#120, T#28) -> [T#156]
Op#21 MUL(T#156, T#129) -> [T#157]
Op#22 MUL(T#153, T#157) -> [T#158]
Op#23 CONV_2D(T#158, T#119, T#71) -> [T#159]
Op#24 CONV_2D(T#159, T#118, T#27) -> [T#160]
Op#25 DEPTHWISE_CONV_2D(T#160, T#70, T#26) -> [T#161]
Op#26 MEAN(T#161, T#130[1, 2]) -> [T#162]
Op#27 CONV_2D(T#162, T#117, T#25) -> [T#163]
Op#28 CONV_2D(T#163, T#116, T#24) -> [T#164]
Op#29 MUL(T#164, T#129) -> [T#165]
Op#30 MUL(T#161, T#165) -> [T#166]
Op#31 CONV_2D(T#166, T#115, T#69) -> [T#167]
Op#32 ADD(T#159, T#167) -> [T#168]
Op#33 CONV_2D(T#168, T#114, T#23) -> [T#169]
Op#34 DEPTHWISE_CONV_2D(T#169, T#68, T#22) -> [T#170]
Op#35 MEAN(T#170, T#130[1, 2]) -> [T#171]
Op#36 CONV_2D(T#171, T#113, T#21) -> [T#172]
Op#37 CONV_2D(T#172, T#112, T#20) -> [T#173]
Op#38 MUL(T#173, T#129) -> [T#174]
Op#39 MUL(T#170, T#174) -> [T#175]
Op#40 CONV_2D(T#175, T#111, T#67) -> [T#176]
Op#41 ADD(T#168, T#176) -> [T#177]
Op#42 CONV_2D(T#177, T#110, T#66) -> [T#178]
Op#43 HARD_SWISH(T#178) -> [T#179]
Op#44 PAD(T#179, T#132[0, 0, 0, 1, 0, ...]) -> [T#180]
Op#45 DEPTHWISE_CONV_2D(T#180, T#65, T#19) -> [T#181]
Op#46 HARD_SWISH(T#181) -> [T#182]
Op#47 CONV_2D(T#182, T#109, T#64) -> [T#183]
Op#48 CONV_2D(T#183, T#108, T#63) -> [T#184]
Op#49 HARD_SWISH(T#184) -> [T#185]
Op#50 DEPTHWISE_CONV_2D(T#185, T#62, T#18) -> [T#186]
Op#51 HARD_SWISH(T#186) -> [T#187]
Op#52 CONV_2D(T#187, T#107, T#61) -> [T#188]
Op#53 ADD(T#183, T#188) -> [T#189]
Op#54 CONV_2D(T#189, T#106, T#60) -> [T#190]
Op#55 HARD_SWISH(T#190) -> [T#191]
Op#56 DEPTHWISE_CONV_2D(T#191, T#59, T#17) -> [T#192]
Op#57 HARD_SWISH(T#192) -> [T#193]
Op#58 CONV_2D(T#193, T#105, T#58) -> [T#194]
Op#59 ADD(T#189, T#194) -> [T#195]
Op#60 CONV_2D(T#195, T#104, T#57) -> [T#196]
Op#61 HARD_SWISH(T#196) -> [T#197]
Op#62 DEPTHWISE_CONV_2D(T#197, T#56, T#16) -> [T#198]
Op#63 HARD_SWISH(T#198) -> [T#199]
Op#64 CONV_2D(T#199, T#103, T#55) -> [T#200]
Op#65 ADD(T#195, T#200) -> [T#201]
Op#66 CONV_2D(T#201, T#102, T#54) -> [T#202]
Op#67 HARD_SWISH(T#202) -> [T#203]
Op#68 DEPTHWISE_CONV_2D(T#203, T#53, T#15) -> [T#204]
Op#69 HARD_SWISH(T#204) -> [T#205]
Op#70 MEAN(T#205, T#130[1, 2]) -> [T#206]
Op#71 CONV_2D(T#206, T#101, T#14) -> [T#207]
Op#72 CONV_2D(T#207, T#100, T#13) -> [T#208]
Op#73 MUL(T#208, T#129) -> [T#209]
Op#74 MUL(T#205, T#209) -> [T#210]
Op#75 CONV_2D(T#210, T#99, T#52) -> [T#211]
Op#76 CONV_2D(T#211, T#98, T#51) -> [T#212]
Op#77 HARD_SWISH(T#212) -> [T#213]
Op#78 DEPTHWISE_CONV_2D(T#213, T#50, T#12) -> [T#214]
Op#79 HARD_SWISH(T#214) -> [T#215]
Op#80 MEAN(T#215, T#130[1, 2]) -> [T#216]
Op#81 CONV_2D(T#216, T#97, T#11) -> [T#217]
Op#82 CONV_2D(T#217, T#96, T#10) -> [T#218]
Op#83 MUL(T#218, T#129) -> [T#219]
Op#84 MUL(T#215, T#219) -> [T#220]
Op#85 CONV_2D(T#220, T#95, T#49) -> [T#221]
Op#86 ADD(T#211, T#221) -> [T#222]
Op#87 CONV_2D(T#222, T#94, T#48) -> [T#223]
Op#88 HARD_SWISH(T#223) -> [T#224]
Op#89 PAD(T#224, T#131[0, 0, 1, 2, 1, ...]) -> [T#225]
Op#90 DEPTHWISE_CONV_2D(T#225, T#47, T#9) -> [T#226]
Op#91 HARD_SWISH(T#226) -> [T#227]
Op#92 MEAN(T#227, T#130[1, 2]) -> [T#228]
Op#93 CONV_2D(T#228, T#93, T#8) -> [T#229]
Op#94 CONV_2D(T#229, T#92, T#7) -> [T#230]
Op#95 MUL(T#230, T#129) -> [T#231]
Op#96 MUL(T#227, T#231) -> [T#232]
Op#97 CONV_2D(T#232, T#91, T#46) -> [T#233]
Op#98 CONV_2D(T#233, T#90, T#45) -> [T#234]
Op#99 HARD_SWISH(T#234) -> [T#235]
Op#100 DEPTHWISE_CONV_2D(T#235, T#44, T#6) -> [T#236]
Op#101 HARD_SWISH(T#236) -> [T#237]
Op#102 MEAN(T#237, T#130[1, 2]) -> [T#238]
Op#103 CONV_2D(T#238, T#89, T#5) -> [T#239]
Op#104 CONV_2D(T#239, T#88, T#4) -> [T#240]
Op#105 MUL(T#240, T#129) -> [T#241]
Op#106 MUL(T#237, T#241) -> [T#242]
Op#107 CONV_2D(T#242, T#87, T#43) -> [T#243]
Op#108 ADD(T#233, T#243) -> [T#244]
Op#109 CONV_2D(T#244, T#86, T#42) -> [T#245]
Op#110 HARD_SWISH(T#245) -> [T#246]
Op#111 DEPTHWISE_CONV_2D(T#246, T#41, T#3) -> [T#247]
Op#112 HARD_SWISH(T#247) -> [T#248]
Op#113 MEAN(T#248, T#130[1, 2]) -> [T#249]
Op#114 CONV_2D(T#249, T#85, T#2) -> [T#250]
Op#115 CONV_2D(T#250, T#84, T#1) -> [T#251]
Op#116 MUL(T#251, T#129) -> [T#252]
Op#117 MUL(T#248, T#252) -> [T#253]
Op#118 CONV_2D(T#253, T#83, T#40) -> [T#254]
Op#119 ADD(T#244, T#254) -> [T#255]
Op#120 CONV_2D(T#255, T#82, T#39) -> [T#256]
Op#121 HARD_SWISH(T#256) -> [T#257]
Op#122 MEAN(T#257, T#130[1, 2]) -> [T#258]
Op#123 CONV_2D(T#258, T#81, T#38) -> [T#259]
Op#124 HARD_SWISH(T#259) -> [T#260]
Op#125 CONV_2D(T#260, T#80, T#37) -> [T#261]
Op#126 RESHAPE(T#261, T#133[-1, 1000]) -> [T#262]
Op#127 SOFTMAX(T#262) -> [T#263]
Tensors of Subgraph#0
T#0(serving_default_keras_tensor_5:0) shape_signature:[-1, -1, -1, 3], type:FLOAT32
T#1(arith.constant) shape:[960], type:FLOAT32 RO 3840 bytes, buffer: 2, data:[0.283771, 0.24407, 0.243922, 1.221, 0.460753, ...]
T#2(arith.constant1) shape:[240], type:FLOAT32 RO 960 bytes, buffer: 3, data:[-0.275301, -0.0277678, -0.411228, -0.3586, -0.220745, ...]
T#3(arith.constant2) shape:[960], type:FLOAT32 RO 3840 bytes, buffer: 4, data:[-1.81109, 1.68503, 1.58476, 1.70023, 0.342517, ...]
T#4(arith.constant3) shape:[960], type:FLOAT32 RO 3840 bytes, buffer: 5, data:[0.195665, 0.217341, 0.114345, -0.0316076, 0.281505, ...]
T#5(arith.constant4) shape:[240], type:FLOAT32 RO 960 bytes, buffer: 6, data:[-0.295283, -0.171183, -0.491539, -0.201764, -0.0582549, ...]
T#6(arith.constant5) shape:[960], type:FLOAT32 RO 3840 bytes, buffer: 7, data:[-1.22443, -0.854031, 1.91604, -3.2009, 0.110498, ...]
T#7(arith.constant6) shape:[672], type:FLOAT32 RO 2688 bytes, buffer: 8, data:[2.06113, 0.736983, 4.40858, 2.36386, 0.687798, ...]
T#8(arith.constant7) shape:[168], type:FLOAT32 RO 672 bytes, buffer: 9, data:[-0.499907, 0.0375283, -0.0576132, -0.243811, -0.391691, ...]
T#9(arith.constant8) shape:[672], type:FLOAT32 RO 2688 bytes, buffer: 10, data:[1.35255, 0.0874219, 0.716237, 0.865584, 1.82332, ...]
T#10(arith.constant9) shape:[672], type:FLOAT32 RO 2688 bytes, buffer: 11, data:[0.291311, 1.62599, 0.179997, 0.249016, 2.76901, ...]
T#11(arith.constant10) shape:[168], type:FLOAT32 RO 672 bytes, buffer: 12, data:[-0.0489284, 0.178251, -0.0412987, -0.205209, 0.0695921, ...]
T#12(arith.constant11) shape:[672], type:FLOAT32 RO 2688 bytes, buffer: 13, data:[-2.30358, -1.0415, -1.02916, -2.42349, -0.143203, ...]
T#13(arith.constant12) shape:[480], type:FLOAT32 RO 1920 bytes, buffer: 14, data:[0.765333, 0.628963, 5.4054, 4.91936, 2.86523, ...]
T#14(arith.constant13) shape:[120], type:FLOAT32 RO 480 bytes, buffer: 15, data:[0.162616, 0.0211225, -0.00731861, 0.275613, 0.465336, ...]
T#15(arith.constant14) shape:[480], type:FLOAT32 RO 1920 bytes, buffer: 16, data:[-1.14594, -1.2222, 0.493229, -0.806949, -0.123236, ...]
T#16(arith.constant15) shape:[184], type:FLOAT32 RO 736 bytes, buffer: 17, data:[-1.82527, -1.90425, -0.864828, -1.20905, 1.78948, ...]
T#17(arith.constant16) shape:[184], type:FLOAT32 RO 736 bytes, buffer: 18, data:[-2.47649, -2.20832, -1.40136, -0.623928, -1.61101, ...]
T#18(arith.constant17) shape:[200], type:FLOAT32 RO 800 bytes, buffer: 19, data:[-1.90742, -1.52078, 4.21308, -1.51046, -1.52174, ...]
T#19(arith.constant18) shape:[240], type:FLOAT32 RO 960 bytes, buffer: 20, data:[2.15248, 1.62511, 4.58976, 2.86807, 1.67084, ...]
T#20(arith.constant19) shape:[120], type:FLOAT32 RO 480 bytes, buffer: 21, data:[1.03397, -0.18951, 3.24036, 1.176, 2.22316, ...]
T#21(arith.constant20) shape:[32], type:FLOAT32 RO 128 bytes, buffer: 22, data:[0.920288, -0.00382053, -0.0567493, 1.97454, 3.35371, ...]
T#22(arith.constant21) shape:[120], type:FLOAT32 RO 480 bytes, buffer: 23, data:[-2.55795, -2.85519, -0.168461, 3.99681, -2.29523, ...]
T#23(arith.constant22) shape:[120], type:FLOAT32 RO 480 bytes, buffer: 24, data:[1.19221, -1.76372, 2.7938, 3.13965, -0.732204, ...]
T#24(arith.constant23) shape:[120], type:FLOAT32 RO 480 bytes, buffer: 25, data:[3.12207, 4.98045, 2.80049, 2.3461, 3.47311, ...]
T#25(arith.constant24) shape:[32], type:FLOAT32 RO 128 bytes, buffer: 26, data:[-0.0122205, 1.39665, 0.193353, 1.20499, -0.000705811, ...]
T#26(arith.constant25) shape:[120], type:FLOAT32 RO 480 bytes, buffer: 27, data:[-0.219226, 0.464636, -0.288737, -2.38097, -0.334142, ...]
T#27(arith.constant26) shape:[120], type:FLOAT32 RO 480 bytes, buffer: 28, data:[3.30378, 2.60396, 2.83121, -4.14912, 2.59554, ...]
T#28(arith.constant27) shape:[72], type:FLOAT32 RO 288 bytes, buffer: 29, data:[5.06759, 6.06202, 5.33617, 6.0275, 4.7227, ...]
T#29(arith.constant28) shape:[24], type:FLOAT32 RO 96 bytes, buffer: 30, data:[1.14102, -0.02167, -0.01928, -0.0118068, 0.218227, ...]
T#30(arith.constant29) shape:[72], type:FLOAT32 RO 288 bytes, buffer: 31, data:[1.70499, 18.0012, 1.05503, 10.0129, -2.74094, ...]
T#31(arith.constant30) shape:[72], type:FLOAT32 RO 288 bytes, buffer: 32, data:[-0.498766, -0.309574, 0.104518, 2.44678, 1.72927, ...]
T#32(arith.constant31) shape:[72], type:FLOAT32 RO 288 bytes, buffer: 33, data:[0.586533, 0.863577, 0.484086, -8.43705, 7.50718, ...]
T#33(arith.constant32) shape:[72], type:FLOAT32 RO 288 bytes, buffer: 34, data:[3.17699, 2.28101, 1.58534, 2.71796, 1.68366, ...]
T#34(arith.constant33) shape:[64], type:FLOAT32 RO 256 bytes, buffer: 35, data:[6.24156, 0.981198, 2.53471, -0.0248699, 25.7691, ...]
T#35(arith.constant34) shape:[64], type:FLOAT32 RO 256 bytes, buffer: 36, data:[5.83326, 7.79689, 5.9951, -0.769312, 8.54113, ...]
T#36(arith.constant35) shape:[16], type:FLOAT32 RO 64 bytes, buffer: 37, data:[1.62813, 33.7453, 4.72859, 8.78206, 17.5393, ...]
T#37(arith.constant36) shape:[1000], type:FLOAT32 RO 4000 bytes, buffer: 38, data:[-0.073695, -0.0658332, -0.00686596, 0.0479387, 0.0198878, ...]
T#38(arith.constant37) shape:[1280], type:FLOAT32 RO 5120 bytes, buffer: 39, data:[0.144575, 0.590702, 0.13199, 0.725449, -0.299175, ...]
T#39(arith.constant38) shape:[960], type:FLOAT32 RO 3840 bytes, buffer: 40, data:[-16.2774, -9.93764, -6.33422, -11.2148, -6.04787, ...]
T#40(arith.constant39) shape:[160], type:FLOAT32 RO 640 bytes, buffer: 41, data:[-1.10576, -6.84556, -0.464385, 3.1173, -3.98359, ...]
T#41(arith.constant40) shape:[1, 5, 5, 960], type:FLOAT32 RO 96000 bytes, buffer: 42, data:[-0.152749, -0.152966, 0.23392, 0.00429554, -0.286706, ...]
T#42(arith.constant41) shape:[960], type:FLOAT32 RO 3840 bytes, buffer: 43, data:[-0.213042, -1.64993, -1.58605, 3.29836, -0.697594, ...]
T#43(arith.constant42) shape:[160], type:FLOAT32 RO 640 bytes, buffer: 44, data:[1.56244, -9.28569, -6.53591, 2.84496, -5.46389, ...]
T#44(arith.constant43) shape:[1, 5, 5, 960], type:FLOAT32 RO 96000 bytes, buffer: 45, data:[4.83438, -1.51938, -0.324659, -0.391306, -0.01447, ...]
T#45(arith.constant44) shape:[960], type:FLOAT32 RO 3840 bytes, buffer: 46, data:[-0.0174746, 0.0162077, -1.22728, 0.279187, -0.554711, ...]
T#46(arith.constant45) shape:[160], type:FLOAT32 RO 640 bytes, buffer: 47, data:[-3.44684, -0.768017, -0.969108, 1.23336, -2.86966, ...]
T#47(arith.constant46) shape:[1, 5, 5, 672], type:FLOAT32 RO 67200 bytes, buffer: 48, data:[-0.00694154, 0.0356305, -0.195693, -0.0262144, 0.114805, ...]
T#48(arith.constant47) shape:[672], type:FLOAT32 RO 2688 bytes, buffer: 49, data:[0.573135, -0.726054, 0.0182186, -0.206486, -1.48872, ...]
T#49(arith.constant48) shape:[112], type:FLOAT32 RO 448 bytes, buffer: 50, data:[0.365523, 1.10257, -1.63187, 0.706468, 0.487061, ...]
T#50(arith.constant49) shape:[1, 3, 3, 672], type:FLOAT32 RO 24192 bytes, buffer: 51, data:[0.0253853, 0.0641128, 1.57708, 0.0533236, -0.00350431, ...]
T#51(arith.constant50) shape:[672], type:FLOAT32 RO 2688 bytes, buffer: 52, data:[2.41973, -1.5073, -0.00963159, -0.640254, 0.684952, ...]
T#52(arith.constant51) shape:[112], type:FLOAT32 RO 448 bytes, buffer: 53, data:[-4.96088, -3.39939, 4.19718, 2.48631, -1.34157, ...]
T#53(arith.constant52) shape:[1, 3, 3, 480], type:FLOAT32 RO 17280 bytes, buffer: 54, data:[-0.0212238, 6.44594, 0.0537825, 0.22657, -0.0316337, ...]
T#54(arith.constant53) shape:[480], type:FLOAT32 RO 1920 bytes, buffer: 55, data:[2.50857, 0.0973693, -0.563608, -1.45203, 3.44066, ...]
T#55(arith.constant54) shape:[80], type:FLOAT32 RO 320 bytes, buffer: 56, data:[-0.435609, -2.97176, 2.74412, -6.65204, 10.2386, ...]
T#56(arith.constant55) shape:[1, 3, 3, 184], type:FLOAT32 RO 6624 bytes, buffer: 57, data:[4.97419, -6.57637, 0.814417, 1.46725, 0.457797, ...]
T#57(arith.constant56) shape:[184], type:FLOAT32 RO 736 bytes, buffer: 58, data:[0.213268, 0.0483445, -0.11253, 0.0761342, -1.73988, ...]
T#58(arith.constant57) shape:[80], type:FLOAT32 RO 320 bytes, buffer: 59, data:[-1.10751, -3.36157, 0.340627, 2.23085, -0.46187, ...]
T#59(arith.constant58) shape:[1, 3, 3, 184], type:FLOAT32 RO 6624 bytes, buffer: 60, data:[0.186774, 0.198745, -0.694211, 0.182543, -0.045065, ...]
T#60(arith.constant59) shape:[184], type:FLOAT32 RO 736 bytes, buffer: 61, data:[1.78543, 1.23138, -0.31343, -2.65884, 2.16531, ...]
T#61(arith.constant60) shape:[80], type:FLOAT32 RO 320 bytes, buffer: 62, data:[2.75995, -0.155923, 2.06222, -4.97617, -12.3297, ...]
T#62(arith.constant61) shape:[1, 3, 3, 200], type:FLOAT32 RO 7200 bytes, buffer: 63, data:[-0.0930532, 1.35916, 0.0699976, 2.08309, -0.714721, ...]
T#63(arith.constant62) shape:[200], type:FLOAT32 RO 800 bytes, buffer: 64, data:[-0.0180047, 0.000351542, 2.84978, 0.00512768, -0.0474478, ...]
T#64(arith.constant63) shape:[80], type:FLOAT32 RO 320 bytes, buffer: 65, data:[-15.6729, 13.1146, 9.85148, 15.7407, -16.4922, ...]
T#65(arith.constant64) shape:[1, 3, 3, 240], type:FLOAT32 RO 8640 bytes, buffer: 66, data:[0.507291, 0.915944, 0.881445, 0.338672, -0.261484, ...]
T#66(arith.constant65) shape:[240], type:FLOAT32 RO 960 bytes, buffer: 67, data:[-3.03785, -3.20833, -1.26339, -0.875435, -0.410649, ...]
T#67(arith.constant66) shape:[40], type:FLOAT32 RO 160 bytes, buffer: 68, data:[-3.93807, 1.26509, -0.947863, 31.8655, 3.26632, ...]
T#68(arith.constant67) shape:[1, 5, 5, 120], type:FLOAT32 RO 12000 bytes, buffer: 69, data:[-0.0563561, -0.887496, 0.0099917, 0.166615, 0.101625, ...]
T#69(arith.constant68) shape:[40], type:FLOAT32 RO 160 bytes, buffer: 70, data:[-7.63688, 1.21586, -22.5861, 0.739685, -3.0402, ...]
T#70(arith.constant69) shape:[1, 5, 5, 120], type:FLOAT32 RO 12000 bytes, buffer: 71, data:[0.0256352, 0.00875715, -0.00830248, 0.0426244, 0.00442088, ...]
T#71(arith.constant70) shape:[40], type:FLOAT32 RO 160 bytes, buffer: 72, data:[-21.1494, -0.469508, 14.1144, -5.10523, -9.47186, ...]
T#72(arith.constant71) shape:[1, 5, 5, 72], type:FLOAT32 RO 7200 bytes, buffer: 73, data:[0.0879386, -0.0954128, 0.0937833, -0.0427546, -0.253503, ...]
T#73(arith.constant72) shape:[24], type:FLOAT32 RO 96 bytes, buffer: 74, data:[-35.7347, 31.8145, 7.77917, 11.8099, 10.6855, ...]
T#74(arith.constant73) shape:[1, 3, 3, 72], type:FLOAT32 RO 2592 bytes, buffer: 75, data:[-0.196813, -0.0441104, -0.806084, 0.0801485, 0.182848, ...]
T#75(arith.constant74) shape:[24], type:FLOAT32 RO 96 bytes, buffer: 76, data:[29.5271, -13.7881, -51.1199, 3.50073, -7.02167, ...]
T#76(arith.constant75) shape:[1, 3, 3, 64], type:FLOAT32 RO 2304 bytes, buffer: 77, data:[-7.61981, 0.609866, -0.72154, 1.24176, -0.446165, ...]
T#77(arith.constant76) shape:[16], type:FLOAT32 RO 64 bytes, buffer: 78, data:[-0.0141129, 49.9822, 9.52096, -9.69061, -4.32951, ...]
T#78(arith.constant77) shape:[1, 3, 3, 16], type:FLOAT32 RO 576 bytes, buffer: 79, data:[1.22061, -0.810988, -0.595521, -0.12323, 0.128769, ...]
T#79(arith.constant78) shape:[16], type:FLOAT32 RO 64 bytes, buffer: 80, data:[26.8229, 27.4359, 2.7004, 6.57344, 25.2757, ...]
T#80(arith.constant79) shape:[1000, 1, 1, 1280], type:FLOAT32 RO 5120000 bytes, buffer: 81, data:[0.0232807, -0.0139387, -0.0507069, 0.0257928, -0.0243703, ...]
T#81(arith.constant80) shape:[1280, 1, 1, 960], type:FLOAT32 RO 4915200 bytes, buffer: 82, data:[0.00633656, -0.0184516, -0.0124931, -0.0359429, 0.00918523, ...]
T#82(arith.constant81) shape:[960, 1, 1, 160], type:FLOAT32 RO 614400 bytes, buffer: 83, data:[0.0500849, -0.0546926, 0.0198143, -0.0325645, -0.219457, ...]
T#83(arith.constant82) shape:[160, 1, 1, 960], type:FLOAT32 RO 614400 bytes, buffer: 84, data:[-2.50493, -3.6528, -3.30272, 0.205339, -1.86695, ...]
T#84(arith.constant83) shape:[960, 1, 1, 240], type:FLOAT32 RO 921600 bytes, buffer: 85, data:[-0.0611927, -0.0151269, -0.0345105, -0.0179798, 0.0131835, ...]
T#85(arith.constant84) shape:[240, 1, 1, 960], type:FLOAT32 RO 921600 bytes, buffer: 86, data:[-0.0837548, -0.0823375, -0.0502755, 0.00375071, 0.0204295, ...]
T#86(arith.constant85) shape:[960, 1, 1, 160], type:FLOAT32 RO 614400 bytes, buffer: 87, data:[-0.00653375, 0.00786634, -0.0076777, 0.00238527, 0.00558404, ...]
T#87(arith.constant86) shape:[160, 1, 1, 960], type:FLOAT32 RO 614400 bytes, buffer: 88, data:[-0.549766, -6.35918, -2.60246, -5.68154, -1.48906, ...]
T#88(arith.constant87) shape:[960, 1, 1, 240], type:FLOAT32 RO 921600 bytes, buffer: 89, data:[-0.000973725, -0.000458092, -0.0380375, -0.00309571, 0.0262516, ...]
T#89(arith.constant88) shape:[240, 1, 1, 960], type:FLOAT32 RO 921600 bytes, buffer: 90, data:[0.0192977, -0.0183088, 0.168897, -0.0208883, -0.0152427, ...]
T#90(arith.constant89) shape:[960, 1, 1, 160], type:FLOAT32 RO 614400 bytes, buffer: 91, data:[0.00076681, -0.000431589, -0.000944783, 0.00120458, 0.00134008, ...]
T#91(arith.constant90) shape:[160, 1, 1, 672], type:FLOAT32 RO 430080 bytes, buffer: 92, data:[2.30253, -1.31009, 0.118996, 1.40242, 1.30476, ...]
T#92(arith.constant91) shape:[672, 1, 1, 168], type:FLOAT32 RO 451584 bytes, buffer: 93, data:[-0.0652855, -0.19256, 0.0154229, -0.0401333, 0.0346401, ...]
T#93(arith.constant92) shape:[168, 1, 1, 672], type:FLOAT32 RO 451584 bytes, buffer: 94, data:[0.0148519, -0.00466832, 0.00745004, 0.00458578, 0.0245794, ...]
T#94(arith.constant93) shape:[672, 1, 1, 112], type:FLOAT32 RO 301056 bytes, buffer: 95, data:[0.0149865, 0.0068462, 0.0173924, -0.00532784, -0.00565932, ...]
T#95(arith.constant94) shape:[112, 1, 1, 672], type:FLOAT32 RO 301056 bytes, buffer: 96, data:[-0.565866, -1.61402, -0.562591, 3.13697, -2.86662, ...]
T#96(arith.constant95) shape:[672, 1, 1, 168], type:FLOAT32 RO 451584 bytes, buffer: 97, data:[0.312729, -0.0241425, 0.0155115, -0.0770605, 0.0287806, ...]
T#97(arith.constant96) shape:[168, 1, 1, 672], type:FLOAT32 RO 451584 bytes, buffer: 98, data:[-0.0403199, 0.0110284, -3.64906e-05, 0.052037, 0.00120152, ...]
T#98(arith.constant97) shape:[672, 1, 1, 112], type:FLOAT32 RO 301056 bytes, buffer: 99, data:[-0.00188285, -0.0111014, -0.044308, -0.0045087, 0.0132006, ...]
T#99(arith.constant98) shape:[112, 1, 1, 480], type:FLOAT32 RO 215040 bytes, buffer: 100, data:[-3.1418, 0.424589, 1.22596, -0.396264, 4.54748, ...]
T#100(arith.constant99) shape:[480, 1, 1, 120], type:FLOAT32 RO 230400 bytes, buffer: 101, data:[-0.120188, 0.0748747, -0.183427, 0.0475327, -0.00263915, ...]
T#101(arith.constant100) shape:[120, 1, 1, 480], type:FLOAT32 RO 230400 bytes, buffer: 102, data:[0.0232848, -0.0156344, 0.0118119, 0.00698492, 0.0173483, ...]
T#102(arith.constant101) shape:[480, 1, 1, 80], type:FLOAT32 RO 153600 bytes, buffer: 103, data:[-0.00826612, 0.0499581, 0.0647706, 0.0257538, -0.00146656, ...]
T#103(arith.constant102) shape:[80, 1, 1, 184], type:FLOAT32 RO 58880 bytes, buffer: 104, data:[0.093478, -0.0599167, 0.0303901, 0.131994, 0.190089, ...]
T#104(arith.constant103) shape:[184, 1, 1, 80], type:FLOAT32 RO 58880 bytes, buffer: 105, data:[-0.0015102, 0.000761898, -0.000109779, 0.000520086, -0.00139291, ...]
T#105(arith.constant104) shape:[80, 1, 1, 184], type:FLOAT32 RO 58880 bytes, buffer: 106, data:[-0.155852, 0.237999, 0.816957, 0.13733, 0.384849, ...]
T#106(arith.constant105) shape:[184, 1, 1, 80], type:FLOAT32 RO 58880 bytes, buffer: 107, data:[-0.000389813, -0.0035757, -0.000356501, -0.00755378, 0.0180794, ...]
T#107(arith.constant106) shape:[80, 1, 1, 200], type:FLOAT32 RO 64000 bytes, buffer: 108, data:[0.306236, -0.0031209, 0.0347371, -0.0932839, 0.142599, ...]
T#108(arith.constant107) shape:[200, 1, 1, 80], type:FLOAT32 RO 64000 bytes, buffer: 109, data:[-0.00231777, -0.000986836, 4.6371e-05, -0.00405917, -0.00202406, ...]
T#109(arith.constant108) shape:[80, 1, 1, 240], type:FLOAT32 RO 76800 bytes, buffer: 110, data:[0.219832, 0.0737946, 0.457842, 0.671469, -0.385924, ...]
T#110(arith.constant109) shape:[240, 1, 1, 40], type:FLOAT32 RO 38400 bytes, buffer: 111, data:[-0.0238429, 0.00749496, 0.0132094, -0.0011158, 0.00737228, ...]
T#111(arith.constant110) shape:[40, 1, 1, 120], type:FLOAT32 RO 19200 bytes, buffer: 112, data:[-0.437176, 0.171119, 0.225141, -0.0630487, 1.52885, ...]
T#112(arith.constant111) shape:[120, 1, 1, 32], type:FLOAT32 RO 15360 bytes, buffer: 113, data:[0.0233749, 1.24947e-32, 1.26503e-32, 0.0765244, -0.00840057, ...]
T#113(arith.constant112) shape:[32, 1, 1, 120], type:FLOAT32 RO 15360 bytes, buffer: 114, data:[0.0396201, 0.0105455, 0.0124103, 0.0153921, 0.0817555, ...]
T#114(arith.constant113) shape:[120, 1, 1, 40], type:FLOAT32 RO 19200 bytes, buffer: 115, data:[0.00788043, -0.00202614, 0.0314182, 0.00642282, 0.0341095, ...]
T#115(arith.constant114) shape:[40, 1, 1, 120], type:FLOAT32 RO 19200 bytes, buffer: 116, data:[-1.25906, 0.497161, -0.0669599, 0.622439, -0.657298, ...]
T#116(arith.constant115) shape:[120, 1, 1, 32], type:FLOAT32 RO 15360 bytes, buffer: 117, data:[-1.24865e-32, 0.0671542, 0.0129759, 0.0901768, -1.7092e-28, ...]
T#117(arith.constant116) shape:[32, 1, 1, 120], type:FLOAT32 RO 15360 bytes, buffer: 118, data:[-1.27762e-32, -1.26314e-32, -1.26755e-32, 1.25007e-32, -1.25637e-32, ...]
T#118(arith.constant117) shape:[120, 1, 1, 40], type:FLOAT32 RO 19200 bytes, buffer: 119, data:[0.0915326, -0.0077282, 0.0208179, 0.013002, -0.0156502, ...]
T#119(arith.constant118) shape:[40, 1, 1, 72], type:FLOAT32 RO 11520 bytes, buffer: 120, data:[0.00945878, -0.0363797, 0.0769495, 0.00295284, 0.0174633, ...]
T#120(arith.constant119) shape:[72, 1, 1, 24], type:FLOAT32 RO 6912 bytes, buffer: 121, data:[0.00195266, 1.24883e-32, 1.24923e-32, 1.2486e-32, 0.000132618, ...]
T#121(arith.constant120) shape:[24, 1, 1, 72], type:FLOAT32 RO 6912 bytes, buffer: 122, data:[-0.00278714, -0.00356496, -0.00289361, -0.00207177, -0.000909253, ...]
T#122(arith.constant121) shape:[72, 1, 1, 24], type:FLOAT32 RO 6912 bytes, buffer: 123, data:[0.0293405, -0.0246265, 0.0406672, -0.019213, 0.0562144, ...]
T#123(arith.constant122) shape:[24, 1, 1, 72], type:FLOAT32 RO 6912 bytes, buffer: 124, data:[-0.291106, 0.053235, 0.518672, -1.19898, 0.418507, ...]
T#124(arith.constant123) shape:[72, 1, 1, 24], type:FLOAT32 RO 6912 bytes, buffer: 125, data:[-0.135531, 0.0593861, -0.00241981, 0.0486889, 0.00179526, ...]
T#125(arith.constant124) shape:[24, 1, 1, 64], type:FLOAT32 RO 6144 bytes, buffer: 126, data:[-0.0101949, 0.105657, -0.0099966, -0.140536, 0.0846852, ...]
T#126(arith.constant125) shape:[64, 1, 1, 16], type:FLOAT32 RO 4096 bytes, buffer: 127, data:[-0.000504434, 0.00025894, -0.000541954, 0.00443714, 0.00453425, ...]
T#127(arith.constant126) shape:[16, 1, 1, 16], type:FLOAT32 RO 1024 bytes, buffer: 128, data:[1.92682e-06, -1.08435e-05, -4.07661e-05, 7.29718e-05, -4.40744e-07, ...]
T#128(arith.constant127) shape:[16, 3, 3, 3], type:FLOAT32 RO 1728 bytes, buffer: 129, data:[2.35286, -1.02746, -1.03095, 3.50268, -1.58557, ...]
T#129(arith.constant128) shape:[], type:FLOAT32 RO 4 bytes, buffer: 130, data:[0.166667]
T#130(arith.constant129) shape:[2], type:INT32 RO 8 bytes, buffer: 131, data:[1, 2]
T#131(arith.constant130) shape:[4, 2], type:INT32 RO 32 bytes, buffer: 132, data:[0, 0, 1, 2, 1, ...]
T#132(arith.constant131) shape:[4, 2], type:INT32 RO 32 bytes, buffer: 133, data:[0, 0, 0, 1, 0, ...]
T#133(arith.constant132) shape:[2], type:INT32 RO 8 bytes, buffer: 134, data:[-1, 1000]
T#134(MobileNetV3Large_1/rescaling_1/Cast/x) shape:[], type:FLOAT32 RO 4 bytes, buffer: 135, data:[0.00784314]
T#135(MobileNetV3Large_1/rescaling_1/Cast_1/x) shape:[], type:FLOAT32 RO 4 bytes, buffer: 136, data:[-1]
T#136(MobileNetV3Large_1/rescaling_1/mul) shape_signature:[-1, -1, -1, 3], type:FLOAT32
T#137(MobileNetV3Large_1/rescaling_1/add) shape_signature:[-1, -1, -1, 3], type:FLOAT32
T#138(MobileNetV3Large_1/conv_bn_1/batchnorm/add_1;MobileNetV3Large_1/conv_1/convolution;) shape_signature:[-1, -1, -1, 16], type:FLOAT32
T#139(MobileNetV3Large_1/activation_1/mul;;MobileNetV3Large_1/activation_1/Relu6;MobileNetV3Large_1/activation_1/add;MobileNetV3Large_1/activation_1/truediv) shape_signature:[-1, -1, -1, 16], type:FLOAT32
T#140(MobileNetV3Large_1/re_lu_1/Relu;MobileNetV3Large_1/expanded_conv_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 16], type:FLOAT32
T#141(MobileNetV3Large_1/expanded_conv_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_project_1/convolution;) shape_signature:[-1, -1, -1, 16], type:FLOAT32
T#142(MobileNetV3Large_1/expanded_conv_add_1/Add) shape_signature:[-1, -1, -1, 16], type:FLOAT32
T#143(MobileNetV3Large_1/re_lu_1_2/Relu;MobileNetV3Large_1/expanded_conv_1_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_1_expand_1/convolution;) shape_signature:[-1, -1, -1, 64], type:FLOAT32
T#144(MobileNetV3Large_1/expanded_conv_1_depthwise_pad_1/Pad) shape_signature:[-1, -1, -1, 64], type:FLOAT32
T#145(MobileNetV3Large_1/re_lu_2_1/Relu;MobileNetV3Large_1/expanded_conv_1_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_1_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_1_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 64], type:FLOAT32
T#146(MobileNetV3Large_1/expanded_conv_1_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_1_project_1/convolution;) shape_signature:[-1, -1, -1, 24], type:FLOAT32
T#147(MobileNetV3Large_1/re_lu_3_1/Relu;MobileNetV3Large_1/expanded_conv_2_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_2_expand_1/convolution;) shape_signature:[-1, -1, -1, 72], type:FLOAT32
T#148(MobileNetV3Large_1/re_lu_4_1/Relu;MobileNetV3Large_1/expanded_conv_2_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_2_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_2_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 72], type:FLOAT32
T#149(MobileNetV3Large_1/expanded_conv_2_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_2_project_1/convolution;) shape_signature:[-1, -1, -1, 24], type:FLOAT32
T#150(MobileNetV3Large_1/expanded_conv_2_add_1/Add) shape_signature:[-1, -1, -1, 24], type:FLOAT32
T#151(MobileNetV3Large_1/re_lu_5_1/Relu;MobileNetV3Large_1/expanded_conv_3_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_3_expand_1/convolution;) shape_signature:[-1, -1, -1, 72], type:FLOAT32
T#152(MobileNetV3Large_1/expanded_conv_3_depthwise_pad_1/Pad) shape_signature:[-1, -1, -1, 72], type:FLOAT32
T#153(MobileNetV3Large_1/re_lu_6_1/Relu;MobileNetV3Large_1/expanded_conv_3_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_3_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_3_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 72], type:FLOAT32
T#154(MobileNetV3Large_1/expanded_conv_3_squeeze_excite_avg_pool_1/Mean) shape_signature:[-1, 1, 1, 72], type:FLOAT32
T#155(MobileNetV3Large_1/expanded_conv_3_squeeze_excite_relu_1/Relu;MobileNetV3Large_1/expanded_conv_3_squeeze_excite_conv_1/add;MobileNetV3Large_1/expanded_conv_3_squeeze_excite_conv_1/convolution;) shape_signature:[-1, 1, 1, 24], type:FLOAT32
T#156(MobileNetV3Large_1/re_lu_7_1/Relu6;MobileNetV3Large_1/Add;MobileNetV3Large_1/expanded_conv_3_squeeze_excite_conv_1_2/add;MobileNetV3Large_1/expanded_conv_3_squeeze_excite_conv_1_2/convolution;) shape_signature:[-1, 1, 1, 72], type:FLOAT32
T#157(MobileNetV3Large_1/Mul) shape_signature:[-1, 1, 1, 72], type:FLOAT32
T#158(MobileNetV3Large_1/expanded_conv_3_squeeze_excite_mul_1/Mul) shape_signature:[-1, -1, -1, 72], type:FLOAT32
T#159(MobileNetV3Large_1/expanded_conv_3_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_3_project_1/convolution;) shape_signature:[-1, -1, -1, 40], type:FLOAT32
T#160(MobileNetV3Large_1/re_lu_8_1/Relu;MobileNetV3Large_1/expanded_conv_4_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_4_expand_1/convolution;) shape_signature:[-1, -1, -1, 120], type:FLOAT32
T#161(MobileNetV3Large_1/re_lu_9_1/Relu;MobileNetV3Large_1/expanded_conv_4_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_4_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_4_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 120], type:FLOAT32
T#162(MobileNetV3Large_1/expanded_conv_4_squeeze_excite_avg_pool_1/Mean) shape_signature:[-1, 1, 1, 120], type:FLOAT32
T#163(MobileNetV3Large_1/expanded_conv_4_squeeze_excite_relu_1/Relu;MobileNetV3Large_1/expanded_conv_4_squeeze_excite_conv_1/add;MobileNetV3Large_1/expanded_conv_4_squeeze_excite_conv_1/convolution;) shape_signature:[-1, 1, 1, 32], type:FLOAT32
T#164(MobileNetV3Large_1/re_lu_10_1/Relu6;MobileNetV3Large_1/Add_1;MobileNetV3Large_1/expanded_conv_4_squeeze_excite_conv_1_2/add;MobileNetV3Large_1/expanded_conv_4_squeeze_excite_conv_1_2/convolution;) shape_signature:[-1, 1, 1, 120], type:FLOAT32
T#165(MobileNetV3Large_1/Mul_1) shape_signature:[-1, 1, 1, 120], type:FLOAT32
T#166(MobileNetV3Large_1/expanded_conv_4_squeeze_excite_mul_1/Mul) shape_signature:[-1, -1, -1, 120], type:FLOAT32
T#167(MobileNetV3Large_1/expanded_conv_4_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_4_project_1/convolution;) shape_signature:[-1, -1, -1, 40], type:FLOAT32
T#168(MobileNetV3Large_1/expanded_conv_4_add_1/Add) shape_signature:[-1, -1, -1, 40], type:FLOAT32
T#169(MobileNetV3Large_1/re_lu_11_1/Relu;MobileNetV3Large_1/expanded_conv_5_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_5_expand_1/convolution;) shape_signature:[-1, -1, -1, 120], type:FLOAT32
T#170(MobileNetV3Large_1/re_lu_12_1/Relu;MobileNetV3Large_1/expanded_conv_5_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_5_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_5_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 120], type:FLOAT32
T#171(MobileNetV3Large_1/expanded_conv_5_squeeze_excite_avg_pool_1/Mean) shape_signature:[-1, 1, 1, 120], type:FLOAT32
T#172(MobileNetV3Large_1/expanded_conv_5_squeeze_excite_relu_1/Relu;MobileNetV3Large_1/expanded_conv_5_squeeze_excite_conv_1/add;MobileNetV3Large_1/expanded_conv_5_squeeze_excite_conv_1/convolution;) shape_signature:[-1, 1, 1, 32], type:FLOAT32
T#173(MobileNetV3Large_1/re_lu_13_1/Relu6;MobileNetV3Large_1/Add_2;MobileNetV3Large_1/expanded_conv_5_squeeze_excite_conv_1_2/add;MobileNetV3Large_1/expanded_conv_5_squeeze_excite_conv_1_2/convolution;) shape_signature:[-1, 1, 1, 120], type:FLOAT32
T#174(MobileNetV3Large_1/Mul_2) shape_signature:[-1, 1, 1, 120], type:FLOAT32
T#175(MobileNetV3Large_1/expanded_conv_5_squeeze_excite_mul_1/Mul) shape_signature:[-1, -1, -1, 120], type:FLOAT32
T#176(MobileNetV3Large_1/expanded_conv_5_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_5_project_1/convolution;) shape_signature:[-1, -1, -1, 40], type:FLOAT32
T#177(MobileNetV3Large_1/expanded_conv_5_add_1/Add) shape_signature:[-1, -1, -1, 40], type:FLOAT32
T#178(MobileNetV3Large_1/expanded_conv_6_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_6_expand_1/convolution;) shape_signature:[-1, -1, -1, 240], type:FLOAT32
T#179(MobileNetV3Large_1/activation_1_2/mul;;MobileNetV3Large_1/activation_1_2/Relu6;MobileNetV3Large_1/activation_1_2/add;MobileNetV3Large_1/activation_1_2/truediv) shape_signature:[-1, -1, -1, 240], type:FLOAT32
T#180(MobileNetV3Large_1/expanded_conv_6_depthwise_pad_1/Pad) shape_signature:[-1, -1, -1, 240], type:FLOAT32
T#181(MobileNetV3Large_1/expanded_conv_6_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_6_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_6_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 240], type:FLOAT32
T#182(MobileNetV3Large_1/activation_2_1/mul;;MobileNetV3Large_1/activation_2_1/Relu6;MobileNetV3Large_1/activation_2_1/add;MobileNetV3Large_1/activation_2_1/truediv) shape_signature:[-1, -1, -1, 240], type:FLOAT32
T#183(MobileNetV3Large_1/expanded_conv_6_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_6_project_1/convolution;) shape_signature:[-1, -1, -1, 80], type:FLOAT32
T#184(MobileNetV3Large_1/expanded_conv_7_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_7_expand_1/convolution;) shape_signature:[-1, -1, -1, 200], type:FLOAT32
T#185(MobileNetV3Large_1/activation_3_1/mul;;MobileNetV3Large_1/activation_3_1/Relu6;MobileNetV3Large_1/activation_3_1/add;MobileNetV3Large_1/activation_3_1/truediv) shape_signature:[-1, -1, -1, 200], type:FLOAT32
T#186(MobileNetV3Large_1/expanded_conv_7_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_7_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_7_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 200], type:FLOAT32
T#187(MobileNetV3Large_1/activation_4_1/mul;;MobileNetV3Large_1/activation_4_1/Relu6;MobileNetV3Large_1/activation_4_1/add;MobileNetV3Large_1/activation_4_1/truediv) shape_signature:[-1, -1, -1, 200], type:FLOAT32
T#188(MobileNetV3Large_1/expanded_conv_7_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_7_project_1/convolution;) shape_signature:[-1, -1, -1, 80], type:FLOAT32
T#189(MobileNetV3Large_1/expanded_conv_7_add_1/Add) shape_signature:[-1, -1, -1, 80], type:FLOAT32
T#190(MobileNetV3Large_1/expanded_conv_8_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_8_expand_1/convolution;) shape_signature:[-1, -1, -1, 184], type:FLOAT32
T#191(MobileNetV3Large_1/activation_5_1/mul;;MobileNetV3Large_1/activation_5_1/Relu6;MobileNetV3Large_1/activation_5_1/add;MobileNetV3Large_1/activation_5_1/truediv) shape_signature:[-1, -1, -1, 184], type:FLOAT32
T#192(MobileNetV3Large_1/expanded_conv_8_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_8_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_8_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 184], type:FLOAT32
T#193(MobileNetV3Large_1/activation_6_1/mul;;MobileNetV3Large_1/activation_6_1/Relu6;MobileNetV3Large_1/activation_6_1/add;MobileNetV3Large_1/activation_6_1/truediv) shape_signature:[-1, -1, -1, 184], type:FLOAT32
T#194(MobileNetV3Large_1/expanded_conv_8_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_8_project_1/convolution;) shape_signature:[-1, -1, -1, 80], type:FLOAT32
T#195(MobileNetV3Large_1/expanded_conv_8_add_1/Add) shape_signature:[-1, -1, -1, 80], type:FLOAT32
T#196(MobileNetV3Large_1/expanded_conv_9_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_9_expand_1/convolution;) shape_signature:[-1, -1, -1, 184], type:FLOAT32
T#197(MobileNetV3Large_1/activation_7_1/mul;;MobileNetV3Large_1/activation_7_1/Relu6;MobileNetV3Large_1/activation_7_1/add;MobileNetV3Large_1/activation_7_1/truediv) shape_signature:[-1, -1, -1, 184], type:FLOAT32
T#198(MobileNetV3Large_1/expanded_conv_9_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_9_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_9_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 184], type:FLOAT32
T#199(MobileNetV3Large_1/activation_8_1/mul;;MobileNetV3Large_1/activation_8_1/Relu6;MobileNetV3Large_1/activation_8_1/add;MobileNetV3Large_1/activation_8_1/truediv) shape_signature:[-1, -1, -1, 184], type:FLOAT32
T#200(MobileNetV3Large_1/expanded_conv_9_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_9_project_1/convolution;) shape_signature:[-1, -1, -1, 80], type:FLOAT32
T#201(MobileNetV3Large_1/expanded_conv_9_add_1/Add) shape_signature:[-1, -1, -1, 80], type:FLOAT32
T#202(MobileNetV3Large_1/expanded_conv_10_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_10_expand_1/convolution;) shape_signature:[-1, -1, -1, 480], type:FLOAT32
T#203(MobileNetV3Large_1/activation_9_1/mul;;MobileNetV3Large_1/activation_9_1/Relu6;MobileNetV3Large_1/activation_9_1/add;MobileNetV3Large_1/activation_9_1/truediv) shape_signature:[-1, -1, -1, 480], type:FLOAT32
T#204(MobileNetV3Large_1/expanded_conv_10_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_10_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_10_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 480], type:FLOAT32
T#205(MobileNetV3Large_1/activation_10_1/mul;;MobileNetV3Large_1/activation_10_1/Relu6;MobileNetV3Large_1/activation_10_1/add;MobileNetV3Large_1/activation_10_1/truediv) shape_signature:[-1, -1, -1, 480], type:FLOAT32
T#206(MobileNetV3Large_1/expanded_conv_10_squeeze_excite_avg_pool_1/Mean) shape_signature:[-1, 1, 1, 480], type:FLOAT32
T#207(MobileNetV3Large_1/expanded_conv_10_squeeze_excite_relu_1/Relu;MobileNetV3Large_1/expanded_conv_10_squeeze_excite_conv_1/add;MobileNetV3Large_1/expanded_conv_10_squeeze_excite_conv_1/convolution;) shape_signature:[-1, 1, 1, 120], type:FLOAT32
T#208(MobileNetV3Large_1/re_lu_14_1/Relu6;MobileNetV3Large_1/Add_3;MobileNetV3Large_1/expanded_conv_10_squeeze_excite_conv_1_2/add;MobileNetV3Large_1/expanded_conv_10_squeeze_excite_conv_1_2/convolution;) shape_signature:[-1, 1, 1, 480], type:FLOAT32
T#209(MobileNetV3Large_1/Mul_3) shape_signature:[-1, 1, 1, 480], type:FLOAT32
T#210(MobileNetV3Large_1/expanded_conv_10_squeeze_excite_mul_1/Mul) shape_signature:[-1, -1, -1, 480], type:FLOAT32
T#211(MobileNetV3Large_1/expanded_conv_10_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_10_project_1/convolution;) shape_signature:[-1, -1, -1, 112], type:FLOAT32
T#212(MobileNetV3Large_1/expanded_conv_11_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_11_expand_1/convolution;) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#213(MobileNetV3Large_1/activation_11_1/mul;;MobileNetV3Large_1/activation_11_1/Relu6;MobileNetV3Large_1/activation_11_1/add;MobileNetV3Large_1/activation_11_1/truediv) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#214(MobileNetV3Large_1/expanded_conv_11_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_11_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_11_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#215(MobileNetV3Large_1/activation_12_1/mul;;MobileNetV3Large_1/activation_12_1/Relu6;MobileNetV3Large_1/activation_12_1/add;MobileNetV3Large_1/activation_12_1/truediv) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#216(MobileNetV3Large_1/expanded_conv_11_squeeze_excite_avg_pool_1/Mean) shape_signature:[-1, 1, 1, 672], type:FLOAT32
T#217(MobileNetV3Large_1/expanded_conv_11_squeeze_excite_relu_1/Relu;MobileNetV3Large_1/expanded_conv_11_squeeze_excite_conv_1/add;MobileNetV3Large_1/expanded_conv_11_squeeze_excite_conv_1/convolution;) shape_signature:[-1, 1, 1, 168], type:FLOAT32
T#218(MobileNetV3Large_1/re_lu_15_1/Relu6;MobileNetV3Large_1/Add_4;MobileNetV3Large_1/expanded_conv_11_squeeze_excite_conv_1_2/add;MobileNetV3Large_1/expanded_conv_11_squeeze_excite_conv_1_2/convolution;) shape_signature:[-1, 1, 1, 672], type:FLOAT32
T#219(MobileNetV3Large_1/Mul_4) shape_signature:[-1, 1, 1, 672], type:FLOAT32
T#220(MobileNetV3Large_1/expanded_conv_11_squeeze_excite_mul_1/Mul) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#221(MobileNetV3Large_1/expanded_conv_11_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_11_project_1/convolution;) shape_signature:[-1, -1, -1, 112], type:FLOAT32
T#222(MobileNetV3Large_1/expanded_conv_11_add_1/Add) shape_signature:[-1, -1, -1, 112], type:FLOAT32
T#223(MobileNetV3Large_1/expanded_conv_12_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_12_expand_1/convolution;) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#224(MobileNetV3Large_1/activation_13_1/mul;;MobileNetV3Large_1/activation_13_1/Relu6;MobileNetV3Large_1/activation_13_1/add;MobileNetV3Large_1/activation_13_1/truediv) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#225(MobileNetV3Large_1/expanded_conv_12_depthwise_pad_1/Pad) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#226(MobileNetV3Large_1/expanded_conv_12_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_12_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_12_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#227(MobileNetV3Large_1/activation_14_1/mul;;MobileNetV3Large_1/activation_14_1/Relu6;MobileNetV3Large_1/activation_14_1/add;MobileNetV3Large_1/activation_14_1/truediv) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#228(MobileNetV3Large_1/expanded_conv_12_squeeze_excite_avg_pool_1/Mean) shape_signature:[-1, 1, 1, 672], type:FLOAT32
T#229(MobileNetV3Large_1/expanded_conv_12_squeeze_excite_relu_1/Relu;MobileNetV3Large_1/expanded_conv_12_squeeze_excite_conv_1/add;MobileNetV3Large_1/expanded_conv_12_squeeze_excite_conv_1/convolution;) shape_signature:[-1, 1, 1, 168], type:FLOAT32
T#230(MobileNetV3Large_1/re_lu_16_1/Relu6;MobileNetV3Large_1/Add_5;MobileNetV3Large_1/expanded_conv_12_squeeze_excite_conv_1_2/add;MobileNetV3Large_1/expanded_conv_12_squeeze_excite_conv_1_2/convolution;) shape_signature:[-1, 1, 1, 672], type:FLOAT32
T#231(MobileNetV3Large_1/Mul_5) shape_signature:[-1, 1, 1, 672], type:FLOAT32
T#232(MobileNetV3Large_1/expanded_conv_12_squeeze_excite_mul_1/Mul) shape_signature:[-1, -1, -1, 672], type:FLOAT32
T#233(MobileNetV3Large_1/expanded_conv_12_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_12_project_1/convolution;) shape_signature:[-1, -1, -1, 160], type:FLOAT32
T#234(MobileNetV3Large_1/expanded_conv_13_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_13_expand_1/convolution;) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#235(MobileNetV3Large_1/activation_15_1/mul;;MobileNetV3Large_1/activation_15_1/Relu6;MobileNetV3Large_1/activation_15_1/add;MobileNetV3Large_1/activation_15_1/truediv) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#236(MobileNetV3Large_1/expanded_conv_13_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_13_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_13_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#237(MobileNetV3Large_1/activation_16_1/mul;;MobileNetV3Large_1/activation_16_1/Relu6;MobileNetV3Large_1/activation_16_1/add;MobileNetV3Large_1/activation_16_1/truediv) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#238(MobileNetV3Large_1/expanded_conv_13_squeeze_excite_avg_pool_1/Mean) shape_signature:[-1, 1, 1, 960], type:FLOAT32
T#239(MobileNetV3Large_1/expanded_conv_13_squeeze_excite_relu_1/Relu;MobileNetV3Large_1/expanded_conv_13_squeeze_excite_conv_1/add;MobileNetV3Large_1/expanded_conv_13_squeeze_excite_conv_1/convolution;) shape_signature:[-1, 1, 1, 240], type:FLOAT32
T#240(MobileNetV3Large_1/re_lu_17_1/Relu6;MobileNetV3Large_1/Add_6;MobileNetV3Large_1/expanded_conv_13_squeeze_excite_conv_1_2/add;MobileNetV3Large_1/expanded_conv_13_squeeze_excite_conv_1_2/convolution;) shape_signature:[-1, 1, 1, 960], type:FLOAT32
T#241(MobileNetV3Large_1/Mul_6) shape_signature:[-1, 1, 1, 960], type:FLOAT32
T#242(MobileNetV3Large_1/expanded_conv_13_squeeze_excite_mul_1/Mul) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#243(MobileNetV3Large_1/expanded_conv_13_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_13_project_1/convolution;) shape_signature:[-1, -1, -1, 160], type:FLOAT32
T#244(MobileNetV3Large_1/expanded_conv_13_add_1/Add) shape_signature:[-1, -1, -1, 160], type:FLOAT32
T#245(MobileNetV3Large_1/expanded_conv_14_expand_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_14_expand_1/convolution;) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#246(MobileNetV3Large_1/activation_17_1/mul;;MobileNetV3Large_1/activation_17_1/Relu6;MobileNetV3Large_1/activation_17_1/add;MobileNetV3Large_1/activation_17_1/truediv) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#247(MobileNetV3Large_1/expanded_conv_14_depthwise_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_14_depthwise_bn_1/batchnorm/mul_1;MobileNetV3Large_1/expanded_conv_14_depthwise_1/depthwise;) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#248(MobileNetV3Large_1/activation_18_1/mul;;MobileNetV3Large_1/activation_18_1/Relu6;MobileNetV3Large_1/activation_18_1/add;MobileNetV3Large_1/activation_18_1/truediv) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#249(MobileNetV3Large_1/expanded_conv_14_squeeze_excite_avg_pool_1/Mean) shape_signature:[-1, 1, 1, 960], type:FLOAT32
T#250(MobileNetV3Large_1/expanded_conv_14_squeeze_excite_relu_1/Relu;MobileNetV3Large_1/expanded_conv_14_squeeze_excite_conv_1/add;MobileNetV3Large_1/expanded_conv_14_squeeze_excite_conv_1/convolution;) shape_signature:[-1, 1, 1, 240], type:FLOAT32
T#251(MobileNetV3Large_1/re_lu_18_1/Relu6;MobileNetV3Large_1/Add_7;MobileNetV3Large_1/expanded_conv_14_squeeze_excite_conv_1_2/add;MobileNetV3Large_1/expanded_conv_14_squeeze_excite_conv_1_2/convolution;) shape_signature:[-1, 1, 1, 960], type:FLOAT32
T#252(MobileNetV3Large_1/Mul_7) shape_signature:[-1, 1, 1, 960], type:FLOAT32
T#253(MobileNetV3Large_1/expanded_conv_14_squeeze_excite_mul_1/Mul) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#254(MobileNetV3Large_1/expanded_conv_14_project_bn_1/batchnorm/add_1;MobileNetV3Large_1/expanded_conv_14_project_1/convolution;) shape_signature:[-1, -1, -1, 160], type:FLOAT32
T#255(MobileNetV3Large_1/expanded_conv_14_add_1/Add) shape_signature:[-1, -1, -1, 160], type:FLOAT32
T#256(MobileNetV3Large_1/conv_1_bn_1/batchnorm/add_1;MobileNetV3Large_1/conv_1_2/convolution;) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#257(MobileNetV3Large_1/activation_19_1/mul;;MobileNetV3Large_1/activation_19_1/Relu6;MobileNetV3Large_1/activation_19_1/add;MobileNetV3Large_1/activation_19_1/truediv) shape_signature:[-1, -1, -1, 960], type:FLOAT32
T#258(MobileNetV3Large_1/global_average_pooling2d_1/Mean) shape_signature:[-1, 1, 1, 960], type:FLOAT32
T#259(MobileNetV3Large_1/conv_2_1/add;MobileNetV3Large_1/conv_2_1/convolution;) shape_signature:[-1, 1, 1, 1280], type:FLOAT32
T#260(MobileNetV3Large_1/activation_20_1/mul;;MobileNetV3Large_1/activation_20_1/Relu6;MobileNetV3Large_1/activation_20_1/add;MobileNetV3Large_1/activation_20_1/truediv) shape_signature:[-1, 1, 1, 1280], type:FLOAT32
T#261(MobileNetV3Large_1/logits_1/add;MobileNetV3Large_1/logits_1/convolution;) shape_signature:[-1, 1, 1, 1000], type:FLOAT32
T#262(MobileNetV3Large_1/flatten_1_1/Reshape) shape_signature:[-1, 1000], type:FLOAT32
T#263(StatefulPartitionedCall_1:0) shape_signature:[-1, 1000], type:FLOAT32
---------------------------------------------------------------
Your TFLite model has '1' signature_def(s).
Signature#0 key: 'serving_default'
- Subgraph: Subgraph#0
- Inputs:
'keras_tensor_5' : T#0
- Outputs:
'output_0' : T#263
---------------------------------------------------------------
Model size: 21930336 bytes
Non-data buffer size: 46804 bytes (00.21 %)
Total data buffer size: 21883532 bytes (99.79 %)
(Zero value buffers): 0 bytes (00.00 %)
* Buffers of TFLite model are mostly used for constant tensors.
And zero value buffers are buffers filled with zeros.
Non-data buffers area are used to store operators, subgraphs and etc.
You can find more details from https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/schema/schema.fbs
Check GPU delegate compatibility
The ModelAnalyzer API provides a way to check the GPU delegate compatibility of the given model by providing gpu_compatibility=True option.
Case 1: When model is incompatibile
The following code shows a way to use gpu_compatibility=True option for simple tf.function which uses tf.slice with a 2D tensor and tf.cosh which are not compatible with GPU delegate.
You will see GPU COMPATIBILITY WARNING per every node which has compatibility issue(s).
import tensorflow as tf
@tf.function(input_signature=[
tf.TensorSpec(shape=[4, 4], dtype=tf.float32)
])
def func(x):
return tf.cosh(x) + tf.slice(x, [1, 1], [1, 1])
converter = tf.lite.TFLiteConverter.from_concrete_functions(
[func.get_concrete_function()], func)
converter.target_spec.supported_ops = [
tf.lite.OpsSet.TFLITE_BUILTINS,
tf.lite.OpsSet.SELECT_TF_OPS,
]
fb_model = converter.convert()
tf.lite.experimental.Analyzer.analyze(model_content=fb_model, gpu_compatibility=True)
=== TFLite ModelAnalyzer ===
Your TFLite model has '1' subgraph(s). In the subgraph description below,
T# represents the Tensor numbers. For example, in Subgraph#0, the FlexCosh op takes
tensor #0 as input and produces tensor #2 as output.
Subgraph#0 main(T#0) -> [T#4]
Op#0 FlexCosh(T#0) -> [T#2]
GPU COMPATIBILITY WARNING: Not supported custom op FlexCosh
Op#1 SLICE(T#0, T#1[1, 1], T#1[1, 1]) -> [T#3]
GPU COMPATIBILITY WARNING: SLICE supports for 3 or 4 dimensional tensors only, but node has 2 dimensional tensors.
Op#2 ADD(T#2, T#3) -> [T#4]
GPU COMPATIBILITY WARNING: Subgraph#0 has GPU delegate compatibility issues at nodes 0, 1 on TFLite runtime version 2.17.0
Tensors of Subgraph#0
T#0(x) shape:[4, 4], type:FLOAT32
T#1(arith.constant) shape:[2], type:INT32 RO 8 bytes, buffer: 2, data:[1, 1]
T#2(Cosh) shape:[4, 4], type:FLOAT32
T#3(Slice) shape:[1, 1], type:FLOAT32
T#4(Identity) shape:[4, 4], type:FLOAT32
---------------------------------------------------------------
Model size: 1140 bytes
Non-data buffer size: 1012 bytes (88.77 %)
Total data buffer size: 128 bytes (11.23 %)
(Zero value buffers): 0 bytes (00.00 %)
* Buffers of TFLite model are mostly used for constant tensors.
And zero value buffers are buffers filled with zeros.
Non-data buffers area are used to store operators, subgraphs and etc.
You can find more details from https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/schema/schema.fbs
I0000 00:00:1721388466.730413 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.733070 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.735514 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.738015 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.740435 13762 devices.cc:67] Number of eligible GPUs (core count >= 8, compute capability >= 0.0): 4
I0000 00:00:1721388466.740882 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.742929 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.744821 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.746860 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.748802 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.750807 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.752686 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.754681 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.756641 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.758657 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.760555 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.762533 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.764618 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.766693 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.768614 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.770586 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.772545 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.774566 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.776509 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721388466.778530 13762 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
W0000 00:00:1721388466.804019 13762 tf_tfl_flatbuffer_helpers.cc:392] Ignored output_format.
W0000 00:00:1721388466.804047 13762 tf_tfl_flatbuffer_helpers.cc:395] Ignored drop_control_dependency.
2024-07-19 11:27:46.820385: W tensorflow/compiler/mlir/lite/flatbuffer_export.cc:3463] TFLite interpreter needs to link Flex delegate in order to run the model since it contains the following Select TFop(s):
Flex ops: FlexCosh
Details:
tf.Cosh(tensor<4x4xf32>) -> (tensor<4x4xf32>) : {device = ""}
See instructions: https://www.tensorflow.org/lite/guide/ops_select
Case 2: When model is compatibile
In this example, the given model is compatbile with GPU delegate.
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(128, 128)),
tf.keras.layers.Dense(256, activation='relu'),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(10)
])
fb_model = tf.lite.TFLiteConverter.from_keras_model(model).convert()
tf.lite.experimental.Analyzer.analyze(model_content=fb_model, gpu_compatibility=True)
/tmpfs/src/tf_docs_env/lib/python3.9/site-packages/keras/src/layers/reshaping/flatten.py:37: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
super().__init__(**kwargs)
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp62e94n92/assets
INFO:tensorflow:Assets written to: /tmpfs/tmp/tmp62e94n92/assets
Saved artifact at '/tmpfs/tmp/tmp62e94n92'. The following endpoints are available:
* Endpoint 'serve'
args_0 (POSITIONAL_ONLY): TensorSpec(shape=(None, 128, 128), dtype=tf.float32, name='keras_tensor_215')
Output Type:
TensorSpec(shape=(None, 10), dtype=tf.float32, name=None)
Captures:
140401615778912: TensorSpec(shape=(), dtype=tf.resource, name=None)
140401615779616: TensorSpec(shape=(), dtype=tf.resource, name=None)
140401615780672: TensorSpec(shape=(), dtype=tf.resource, name=None)
140406936389056: TensorSpec(shape=(), dtype=tf.resource, name=None)
W0000 00:00:1721388467.162677 13762 tf_tfl_flatbuffer_helpers.cc:392] Ignored output_format.
W0000 00:00:1721388467.162703 13762 tf_tfl_flatbuffer_helpers.cc:395] Ignored drop_control_dependency.
=== TFLite ModelAnalyzer ===
Your TFLite model has '1' subgraph(s). In the subgraph description below,
T# represents the Tensor numbers. For example, in Subgraph#0, the RESHAPE op takes
tensor #0 and tensor #3 as input and produces tensor #4 as output.
Subgraph#0 main(T#0) -> [T#6]
Op#0 RESHAPE(T#0, T#3[-1, 16384]) -> [T#4]
Op#1 FULLY_CONNECTED(T#4, T#2, T#-1) -> [T#5]
Op#2 FULLY_CONNECTED(T#5, T#1, T#-1) -> [T#6]
Tensors of Subgraph#0
T#0(serving_default_keras_tensor_215:0) shape_signature:[-1, 128, 128], type:FLOAT32
T#1(arith.constant) shape:[10, 256], type:FLOAT32 RO 10240 bytes, buffer: 2, data:[0.0262714, -0.00773203, -0.0964963, -0.0102322, -0.0330515, ...]
T#2(arith.constant1) shape:[256, 16384], type:FLOAT32 RO 16777216 bytes, buffer: 3, data:[-0.00830283, -0.000614734, 0.00744855, 0.00309602, -0.00465189, ...]
T#3(arith.constant2) shape:[2], type:INT32 RO 8 bytes, buffer: 4, data:[-1, 16384]
T#4(sequential_1_1/flatten_2_1/Reshape) shape_signature:[-1, 16384], type:FLOAT32
T#5(sequential_1_1/dense_2_1/MatMul;sequential_1_1/dense_2_1/Relu;sequential_1_1/dense_2_1/Add) shape_signature:[-1, 256], type:FLOAT32
T#6(StatefulPartitionedCall_1:0) shape_signature:[-1, 10], type:FLOAT32
Your model looks compatible with GPU delegate on TFLite runtime version 2.17.0.
This does not guarantee that your model will work well with GPU delegate because there could still be runtime incompatibililties.
---------------------------------------------------------------
Your TFLite model has '1' signature_def(s).
Signature#0 key: 'serving_default'
- Subgraph: Subgraph#0
- Inputs:
'keras_tensor_215' : T#0
- Outputs:
'output_0' : T#6
---------------------------------------------------------------
Model size: 16789056 bytes
Non-data buffer size: 1480 bytes (00.01 %)
Total data buffer size: 16787576 bytes (99.99 %)
(Zero value buffers): 0 bytes (00.00 %)
* Buffers of TFLite model are mostly used for constant tensors.
And zero value buffers are buffers filled with zeros.
Non-data buffers area are used to store operators, subgraphs and etc.
You can find more details from https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/schema/schema.fbs