sushreebarsa

icon location Bangalore

Results 267 comments of sushreebarsa

`Square` and `Sqrt` ops do not support complex values on GPU

@sachinprasadhs I was able to reproduce this issue on colab , please find the gist [here](https://colab.research.google.com/gist/sushreebarsa/dd05c8de1338abea328d9ca18b9f3c99/square_sqrt_ops.ipynb#scrollTo=eiaunmDpLjTA).Thanks!

backing_device changes depending on the tensor's dtype

Was able to reproduce the issue in TF 2.6.0-dev20210528,please find the gist [here](https://colab.research.google.com/gist/sushreebarsa/cbcca6a474a1a37a3f09fc52cbabf24a/untitled49.ipynb#scrollTo=hr6nP5jy59AQ) ..Thanks !

`Cannot iterate over a scalar tensor.` with split_dataset

@mihalt Could you please provide the complete standalone code to replicate this issue? Thank you!

ButchNormalization fails when nvidia GPU is used and the training size and the batch size is "well set"

Was able to reproduce the issue with TF v2.5-gpu,please find the gist [here](https://colab.research.google.com/gist/sushreebarsa/4a89ff3d35fac190ffd204cbe6c1f015/untitled599.ipynb#scrollTo=mdDdU-pI-J-P) ..Thanks!

different result on dict of tensors by getting `_enable_dict_to_input_mapping` in tf.python.keras.engine.functional.py

Was able to reproduce the issue in TF v2.5,please check the gist [here](https://colab.research.google.com/gist/sushreebarsa/8c51b0cdf32c34e08ee0468239ad3d27/untitled54.ipynb)..Thanks !

Mask that had been removed from a tensor re-appears when building a new model with placeholders

@gadagashwini I was able to replicate this issue on colab using TF v2.8.0 and tf-nightly,please find the gist[ here](https://colab.research.google.com/gist/sushreebarsa/9cf59b97f31758483496dc727ec1e51f/16441.ipynb).Thanks!

Given an infinity input, ConvLSTM2D will output a normal value only when the cols shape is 1

@maybeLee I tried to reproduce the issue with the provided code on colab using TF v2.8.0 and faced a different outcome as below .Could you please find the gist [here](https://colab.research.google.com/gist/sushreebarsa/60f463736218ca5639aeb270f02b4322/16492.ipynb#scrollTo=_s6tggqy3N4q)...

Tflite model fails on Android GPU delegate due to 'null pointer dereference'

@ramonhollands Please ensure that the GPU delegate is configured correctly in your Android application and you are using TF latest version as well? Thank you!

Numerical instability of gradient calculation of tf.norm (nan at 0, inf for small values)

Hi, A common and effective approach is to add a small epsilon value (ε) to the squared sum before taking the square root in the norm calculation. This prevents division...

Data Generator extending keras.utils.Sequence uses index=0 twice at the very first iteration

Hello @KIC! In order to expedite the trouble-shooting process, please provide the complete code snippet to reproduce the issue reported here. Thank you!