#### Martin Molnar

Self-taught learner

Hello! I'm a self-taught machine learning enthusiast, still learning more.

Share

In this notebook, I try to create a new tokenization algorithm for NLP (disclaimer, it didn't go too well), where it's more like a tokenization

*layer*(that trains), and it goes at the front of a neural network trying to solve some NLP task.

Here's the link to the notebook.

Share

Share

Hello, I've been trying to get a newer version of TensorFlow (i.e. version 2.6.0rc0), however I can't get it to work. Here's the command I'm using:

!pip install --upgrade --user tensorflow==2.6.0rc0

This works fine, except for a warning about some scripts not being installed on PATH (pastebin here).

Then, when I try to import tensorflow as tf, I get:

Then, when I try to import tensorflow as tf, I get:

NotFoundError: /root/.local/lib/python3.7/site-packages/tensorflow/core/kernels/libtfkernel_sobol_op.so: undefined symbol: _ZN10tensorflow14kernel_factory17OpKernelRegistrar12InitInternalEPKNS_9KernelDefEN4absl12lts_2021032411string_viewESt10unique_ptrINS0_15OpKernelFactoryESt14default_deleteIS9_EE

How do I get TensorFlow version 2.6.0rc0 in Deepnote? Thanks!

Share

I've just published my first notebook! It concerns a new technique I've came up with for neural networks; when you have new neurons you would like to use, but their gradient can't be traversed by gradient descent, you can employ a "gradient pseudo-swap" 🙂

Here's the link to the notebook.

Any (enactable) feedback is welcomed (and appreciated).

Thanks for your time!

Martin