Tensorflow Plugin🔗

The Tensorflow plugin for vAccel implements basic Tensorflow and Tensorflow Lite support for vAccel operations.

Supported operations🔗

Installing Tensorflow and Tensorflow Lite C/C++ API files🔗

You can find instructions on how to install the required files on the Build and Install Tensorflow page. The rest of this guide assumes Tensorflow is installed at /usr/local and its' libraries are at /usr/local/lib.

Installing the plugin🔗

You can get the latest Tensorflow plugin binary release from the Releases page of the vAccel repository. Releases include binaries for x86_64/aarch64 Ubuntu-based systems.

Requirements🔗

The prebuilt TF plugin binaries depend on libarchive. You can install it with:

sudo apt install libarchive13

TAR🔗

To install the TAR binary package of the latest Tensorflow plugin release:

x86ARM (64-bit)

wget https://github.com/nubificus/vaccel/releases/download/v0.8.0/vaccel-tf_0.3.0+vaccel.0.8.0_x86_64.tar.gz
# Replace '/usr/local' below with the desired installation prefix
tar xfv vaccel-tf_0.3.0+vaccel.0.8.0_x86_64.tar.gz --strip-components=2 -C /usr/local
# Update pkg-config files with the correct prefix
find /usr/local -name "vaccel-tf.pc" -exec sed -i 's:^\(prefix=\).*:\1/usr/local:g' {} \;

wget https://github.com/nubificus/vaccel/releases/download/v0.8.0/vaccel-tf_0.3.0+vaccel.0.8.0_aarch64.tar.gz
# Replace '/usr/local' below with the desired installation prefix
tar xfv vaccel-tf_0.3.0+vaccel.0.8.0_aarch64.tar.gz --strip-components=2 -C /usr/local
# Update pkg-config files with the correct prefix
find /usr/local -name "vaccel-tf.pc" -exec sed -i 's:^\(prefix=\).*:\1/usr/local:g' {} \;

Latest artifacts🔗

You can also find prebuilt artifacts of the latest Tensorflow plugin revision at:

x86ARM (64-bit)

# TAR
https://s3.nbfc.io/nbfc-assets/github/vaccel/plugins/tf/rev/main/x86_64/release/vaccel-tf_latest_x86_64.tar.gz

# TAR
https://s3.nbfc.io/nbfc-assets/github/vaccel/plugins/tf/rev/main/aarch64/release/vaccel-tf_latest_aarch64.tar.gz

Usage🔗

To specify Tensorflow plugin as the selected plugin for vAccel execution:

export VACCEL_PLUGINS=libvaccel-tf.so

Ensure Tensorflow, vAccel and the Tensorflow plugin libraries are in the library search paths before trying to use the plugin.

Running an example🔗

Export the necessary variables:

export VACCEL_PLUGINS=libvaccel-tf.so
# Optionally, for verbose output
export VACCEL_LOG_LEVEL=4

Assuming vAccel is installed at /usr/local, you can run an image classification with a ResNet model from https://s3.nbfc.io/models/tf/resnet18-v2-7_saved_model.tar.xz with:

$ tf_inference \
      /usr/local/share/vaccel/images/example.jpg \
      https://s3.nbfc.io/models/tf/resnet18-v2-7_saved_model.tar.xz \
      /usr/local/share/vaccel/labels/imagenet.txt
2026-04-29 15:13:10.336778: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026-04-29 15:13:10.358481: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026.04.29-15:13:10.26 - <debug> Initializing vAccel
2026.04.29-15:13:10.26 - <info> vAccel 0.7.1-93-ebc23b1f
2026.04.29-15:13:10.26 - <debug> Config:
2026.04.29-15:13:10.26 - <debug>   plugins = libvaccel-tf.so
2026.04.29-15:13:10.26 - <debug>   log_level = debug
2026.04.29-15:13:10.26 - <debug>   log_file = (null)
2026.04.29-15:13:10.26 - <debug>   profiling_enabled = false
2026.04.29-15:13:10.26 - <debug>   version_ignore = false
2026.04.29-15:13:10.26 - <debug> Created top-level rundir: /run/user/0/vaccel/9QDV7g
2026.04.29-15:13:10.36 - <info> Registered plugin tf 0.2.0-22-b9061d01
2026.04.29-15:13:10.36 - <debug> Registered op tf_model_load from plugin tf
2026.04.29-15:13:10.36 - <debug> Registered op tf_model_unload from plugin tf
2026.04.29-15:13:10.36 - <debug> Registered op tf_model_run from plugin tf
2026.04.29-15:13:10.36 - <debug> Registered op tflite_model_load from plugin tf
2026.04.29-15:13:10.36 - <debug> Registered op tflite_model_unload from plugin tf
2026.04.29-15:13:10.36 - <debug> Registered op tflite_model_run from plugin tf
2026.04.29-15:13:10.36 - <debug> Loaded plugin tf from libvaccel-tf.so
2026.04.29-15:13:10.36 - <debug> Initialized resource 1
Initialized model resource 1
2026.04.29-15:13:10.36 - <debug> New rundir for session 1: /run/user/0/vaccel/9QDV7g/session.1
2026.04.29-15:13:10.36 - <debug> Initialized session 1 with plugin tf
Initialized vAccel session 1
2026.04.29-15:13:10.36 - <debug> New rundir for resource 1: /run/user/0/vaccel/9QDV7g/resource.1
2026.04.29-15:13:10.36 - <debug> Downloading https://s3.nbfc.io/models/tf/resnet18-v2-7_saved_model.tar.xz
2026.04.29-15:13:11.40 - <debug> Downloaded: 41.0 MB of 41.0 MB (100.0%) | Speed: 39.37 MB/sec
2026.04.29-15:13:11.40 - <debug> Download completed successfully
2026.04.29-15:13:11.40 - <debug> session:1 Registered resource 1
2026.04.29-15:13:11.40 - <debug> session:1 Looking for func implementing op tf_model_load
2026.04.29-15:13:11.40 - <debug> Returning func for op tf_model_load from plugin tf
2026.04.29-15:13:13.52 - <debug> [tf] Loading session from SavedModel
2026.04.29-15:13:13.53 - <debug> [tf] Extracting: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/
2026.04.29-15:13:13.53 - <debug> [tf] Extracting: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/assets/
2026.04.29-15:13:13.53 - <debug> [tf] Extracting: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/fingerprint.pb
2026.04.29-15:13:13.53 - <debug> [tf] Extracting: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/saved_model.pb
2026.04.29-15:13:15.66 - <debug> [tf] Extracting: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/variables/
2026.04.29-15:13:15.66 - <debug> [tf] Extracting: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/variables/variables.data-00000-of-00001
2026.04.29-15:13:15.66 - <debug> [tf] Extracting: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/variables/variables.index
2026-04-29 15:13:15.669831: I tensorflow/cc/saved_model/reader.cc:83] Reading SavedModel from: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model
2026-04-29 15:13:15.690971: I tensorflow/cc/saved_model/reader.cc:52] Reading meta graph with tags { serve }
2026-04-29 15:13:15.691028: I tensorflow/cc/saved_model/reader.cc:147] Reading SavedModel debug info (if present) from: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model
2026-04-29 15:13:15.691160: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: SSE3 SSE4.1 SSE4.2 AVX AVX2 AVX_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2026-04-29 15:13:15.777876: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:388] MLIR V1 optimization pass is not enabled
2026-04-29 15:13:15.778679: I tensorflow/cc/saved_model/loader.cc:236] Restoring SavedModel bundle.
2026-04-29 15:13:15.815605: I tensorflow/cc/saved_model/loader.cc:220] Running initialization op on SavedModel bundle at path: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model
2026-04-29 15:13:15.849330: I tensorflow/cc/saved_model/loader.cc:462] SavedModel load for tags { serve }; Status: success: OK. Took 179498 microseconds.
2026.04.29-15:13:15.90 - <debug> [tf] Model loaded correctly
Session load status => code:0 message:
2026.04.29-15:13:15.91 - <debug> session:1 Looking for func implementing op tf_model_run
2026.04.29-15:13:15.91 - <debug> Returning func for op tf_model_run from plugin tf
2026.04.29-15:13:15.91 - <debug> [tf] Running session
2026.04.29-15:13:16.65 - <debug> [tf] Success
Session run status => code:0 message:
Success!
Output tensor => type:1 nr_dims:2 size:4000B
Prediction: banana
2026.04.29-15:13:16.65 - <debug> session:1 Looking for func implementing op tf_model_unload
2026.04.29-15:13:16.65 - <debug> Returning func for op tf_model_unload from plugin tf
2026.04.29-15:13:16.65 - <debug> [tf] Deleting session
2026.04.29-15:13:16.66 - <debug> [tf] Removed file: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/fingerprint.pb
2026.04.29-15:13:16.66 - <debug> [tf] Removed file: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/variables/variables.index
2026.04.29-15:13:16.66 - <debug> [tf] Removed file: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/variables/variables.data-00000-of-00001
2026.04.29-15:13:16.66 - <debug> [tf] Removed directory: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/variables
2026.04.29-15:13:16.66 - <debug> [tf] Removed file: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/saved_model.pb
2026.04.29-15:13:16.66 - <debug> [tf] Removed directory: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model/assets
2026.04.29-15:13:16.66 - <debug> [tf] Removed directory: /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model
Session delete status => code:0 message:
2026.04.29-15:13:16.66 - <debug> session:1 Unregistered resource 1
2026.04.29-15:13:16.66 - <debug> Released session 1
2026.04.29-15:13:16.66 - <debug> Removing file /run/user/0/vaccel/9QDV7g/resource.1/resnet18-v2-7_saved_model.tar.xz
2026.04.29-15:13:16.67 - <debug> Released resource 1
2026.04.29-15:13:16.67 - <debug> Cleaning up vAccel
2026.04.29-15:13:16.67 - <debug> Cleaning up sessions
2026.04.29-15:13:16.67 - <debug> Cleaning up resources
2026.04.29-15:13:16.67 - <debug> Cleaning up plugins
2026.04.29-15:13:16.67 - <debug> Unregistered plugin tf

You can also run a similar inference example using Tensorflow Lite with:

$ tflite_inference \
      /usr/local/share/vaccel/images/example.jpg \
      https://s3.nbfc.io/models/tf/resnet18-v2-7_float32.tflite \
      /usr/local/share/vaccel/labels/imagenet.txt
2026-04-29 15:14:03.525122: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026-04-29 15:14:03.546097: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026.04.29-15:14:03.49 - <debug> Initializing vAccel
2026.04.29-15:14:03.49 - <info> vAccel 0.7.1-93-ebc23b1f
2026.04.29-15:14:03.49 - <debug> Config:
2026.04.29-15:14:03.49 - <debug>   plugins = libvaccel-tf.so
2026.04.29-15:14:03.49 - <debug>   log_level = debug
2026.04.29-15:14:03.49 - <debug>   log_file = (null)
2026.04.29-15:14:03.49 - <debug>   profiling_enabled = false
2026.04.29-15:14:03.49 - <debug>   version_ignore = false
2026.04.29-15:14:03.49 - <debug> Created top-level rundir: /run/user/0/vaccel/NFiJEK
2026.04.29-15:14:03.54 - <info> Registered plugin tf 0.2.0-22-b9061d01
2026.04.29-15:14:03.54 - <debug> Registered op tf_model_load from plugin tf
2026.04.29-15:14:03.54 - <debug> Registered op tf_model_unload from plugin tf
2026.04.29-15:14:03.54 - <debug> Registered op tf_model_run from plugin tf
2026.04.29-15:14:03.54 - <debug> Registered op tflite_model_load from plugin tf
2026.04.29-15:14:03.54 - <debug> Registered op tflite_model_unload from plugin tf
2026.04.29-15:14:03.54 - <debug> Registered op tflite_model_run from plugin tf
2026.04.29-15:14:03.54 - <debug> Loaded plugin tf from libvaccel-tf.so
2026.04.29-15:14:03.54 - <debug> Initialized resource 1
Initialized model resource 1
2026.04.29-15:14:03.54 - <debug> New rundir for session 1: /run/user/0/vaccel/NFiJEK/session.1
2026.04.29-15:14:03.54 - <debug> Initialized session 1 with plugin tf
Initialized vAccel session 1
2026.04.29-15:14:03.54 - <debug> New rundir for resource 1: /run/user/0/vaccel/NFiJEK/resource.1
2026.04.29-15:14:03.54 - <debug> Downloading https://s3.nbfc.io/models/tf/resnet18-v2-7_float32.tflite
2026.04.29-15:14:04.73 - <debug> Downloaded: 44.6 MB of 44.6 MB (100.0%) | Speed: 37.63 MB/sec
2026.04.29-15:14:04.73 - <debug> Download completed successfully
2026.04.29-15:14:04.73 - <debug> session:1 Registered resource 1
2026.04.29-15:14:04.73 - <debug> session:1 Looking for func implementing op tflite_model_load
2026.04.29-15:14:04.73 - <debug> Returning func for op tflite_model_load from plugin tf
2026.04.29-15:14:04.73 - <debug> [tf][tflite] Loading session from model
2026.04.29-15:14:04.73 - <debug> [tf][tflite] Model loaded correctly
2026.04.29-15:14:04.74 - <debug> session:1 Looking for func implementing op tflite_model_run
2026.04.29-15:14:04.74 - <debug> Returning func for op tflite_model_run from plugin tf
2026.04.29-15:14:04.74 - <debug> [tf][tflite] Running session
2026.04.29-15:14:04.84 - <debug> [tf][tflite] Success
Session run status: 0
Success!
Output tensor => type:1 nr_dims:2 size:4000B
Prediction: banana
2026.04.29-15:14:04.84 - <debug> session:1 Looking for func implementing op tflite_model_unload
2026.04.29-15:14:04.84 - <debug> Returning func for op tflite_model_unload from plugin tf
2026.04.29-15:14:04.84 - <debug> session:1 Unregistered resource 1
2026.04.29-15:14:04.84 - <debug> Released session 1
2026.04.29-15:14:04.84 - <debug> Removing file /run/user/0/vaccel/NFiJEK/resource.1/resnet18-v2-7_float32.tflite
2026.04.29-15:14:04.85 - <debug> Released resource 1
2026.04.29-15:14:04.85 - <debug> Cleaning up vAccel
2026.04.29-15:14:04.85 - <debug> Cleaning up sessions
2026.04.29-15:14:04.85 - <debug> Cleaning up resources
2026.04.29-15:14:04.85 - <debug> Cleaning up plugins
2026.04.29-15:14:04.85 - <debug> Unregistered plugin tf

Using the vAccel Tensorflow bindings🔗

This plugin includes bindings that implement simple inference operations. The bindings can be used to run native Tensorflow application with vAccel. You can find more information in the relevant page.