Spaces:
Sleeping
Sleeping
chore: add note on the other models of tf.keras.applications.
Browse files
main.py
CHANGED
@@ -23,6 +23,7 @@ As a consequence, you might have to wait for a few minutes to note the results.
|
|
23 |
* This Space makes use of the [Docker x Space integration](https://huggingface.co/docs/hub/spaces-sdks-docker) to perform the TensorRT optimizations.
|
24 |
* The default TensorFlow installation doesn't come loaded with a correctly compiled TensorRT. This is why it's recommended to use an [NVIDIA container](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/tensorflow) to perform your TensorRT-related stuff. This is also why the Docker x Space integration was used in this Space.
|
25 |
* To get the maximum peformance, one must use the same hardware for inference as the one used for running the optimizations. For example, if you used a T4-based machine to perform the optimizations, ensure that you're using the same GPU while running inference with your optimized model.
|
|
|
26 |
* One is encouraged to try out different forms of post-training quantization as shown in [this notebook](https://github.com/tensorflow/tensorrt/blob/master/tftrt/benchmarking-python/image_classification/NGC-TFv2-TF-TRT-inference-from-Keras-saved-model.ipynb) to squeeze out the maximum performance using NVIDIA hardware and TensorRT.
|
27 |
"""
|
28 |
|
|
|
23 |
* This Space makes use of the [Docker x Space integration](https://huggingface.co/docs/hub/spaces-sdks-docker) to perform the TensorRT optimizations.
|
24 |
* The default TensorFlow installation doesn't come loaded with a correctly compiled TensorRT. This is why it's recommended to use an [NVIDIA container](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/tensorflow) to perform your TensorRT-related stuff. This is also why the Docker x Space integration was used in this Space.
|
25 |
* To get the maximum peformance, one must use the same hardware for inference as the one used for running the optimizations. For example, if you used a T4-based machine to perform the optimizations, ensure that you're using the same GPU while running inference with your optimized model.
|
26 |
+
* One can use this Space to optimize the others models provided in [tf.keras.applications](https://keras.io/api/applications/).
|
27 |
* One is encouraged to try out different forms of post-training quantization as shown in [this notebook](https://github.com/tensorflow/tensorrt/blob/master/tftrt/benchmarking-python/image_classification/NGC-TFv2-TF-TRT-inference-from-Keras-saved-model.ipynb) to squeeze out the maximum performance using NVIDIA hardware and TensorRT.
|
28 |
"""
|
29 |
|