-
### 1. System information
- OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
- TensorFlow installation (pip package or built from source):
- TensorFlow library (version, if pip package or…
-
For consuming various data formats (mp3 files in my case), I'd like to bundle relevant decoding ops in my SavedModel so there's no need for orchestrating additional frontend services with particular I…
-
## Is your feature request related to a problem? Please describe.
- Currently to implement ensemble of models, the user needs to create a pipeline and manage data flow on the client side.
- This int…
-
Release Date: mid november 2021 ~~october~~ ~~september~~
Wanted:
- Infrastructure:
- [x] PyPy3.7-7.3.7 (with 3 weeks long regressions finally hunted down)
- [x] PyPy3.8-7.3.7 beta,…
-
Dear Nvidia Team,
I am trying to Serve the Custom Model using Triton Server on my PC(having RTX 2060 GPU).
I am using nvcr.io/nvidia/tensorrtserver:20.07-py3 container for Serving.
While I am s…
-
Most of the time I want to save the best models instead of the most recent models. Doing so using tf.train.Saver requires to choose when to save a model and to delete the worst model (which might not …
-
I want the outputs from tfserving is: `{"outputs":{"scores":[0.936071157,0.0527787767]}}`
but I got : `{ "outputs": [ 0.936071038, 0.0527787916 ] }`
Environment:
python 3.6
tensorflow 2.3.0
tf…
-
### 1. System information
- Google Colab
-Tensorflow 2
### 2. Code
https://github.com/Akhilesh64/Image-Segmentation-U-2-Net/blob/main/U2-net.py
#### Option A: Reference code
1) Reference …
-
Support tflite conversion and serving
-
Hello,
Running main.cpp for efficientnet throws this error on the line that uses the cppflow::arg_max(output,1). What could be the cause here?
"terminate called after throwing an instance of 's…