BrainScript | A BrainF * * * interpreter written in TypeScript | Interpreter library

As asked in the title, i would like to know if it is possible to make a model early-stop the epochs during training when the error is reduced enough, so i can avoid overfitting and guessing the right number of epochs at each call.

This is the only thing i have found in the official documentation but it is tobe used in brainscript, and i don't know a single thing about it. I'm using Python 3.6 with CNTK 2.6.

Also, is there a way to perform cross validation in a CNTK CNN?? How could this be done?

Thanks in advance.

This seems to be possible using the Python and C# APIs as per the documentation.

The BrainScript documentation doesn't have Adam as one of the options for gradUpdateType.

Is it possible to use the Adam optimizer in BrainScript?

I have an N-Series Azure VM (the Data Science VM) with Tesla K80 GPU. According to the NVIDIA scanner my GPU driver is up to date. When I run my CNTK Brainscript it says "No GPUs Found" and runs in CPU mode. What can I do to troubleshoot?

CNTK Brainscript provides function ReconcileDynamicAxis for dealing with multiple dynamic axes (e.g. multiple sequence inputs of different size). I didn't find any official documentation about similar Python API. Finally, following some code available on CNTK github, I figured out a way to use it:

from _cntk_py import reconcile_dynamic_axis [...] reconciled_vector_A = typemap(reconcile_dynamic_axis)(vector_A, vector_B)

...it seems to work, but it looks like it's some internal CNTK function and possibly wasn't intended for public use.

Is there a public API version of reconcile_dynamic_axis and if not, what is the recommended way of dealing with multiple dynamic axes in Python API?

I have trained my model with BrainScript and now want to use it in a C++ project. From the EvalMultithreads.cpp I copied all the relevant functions. I linked the CNTK.Core and CNTK.Math library and set their search path as described here. Of course I included the CNTKLibrary.h. When I build the project the compiler gives me this error:

undefined reference to `CNTK::Variable::Name[abi:cxx11]() const'

It referrers to this line :

I'm migrating training and evaluation configurations for CNTK from Brainscript over to Python. Because our training data is created in a map/reduce framework, I'm storing images and labels in a huge text file that contains the base64 encoded image as one of its columns. That all worked fine, but I still have not found a way of doing the equivalent in Python.

My CNTK.exe configuration is similar to this example configuration:

What are the recommendations for creating, saving, loading and evaluating models with multiple outputs in CNTK Python APIs? (I'm using CNTK 2.0 RC1)

When I define a network in run-time, I can evaluate any layer of the network easily, by passing a minibatch dictionary to the eval function.

However, after I save the model and load it again, I can evaluate only the output node by providing the inputs (and I don't know how to define multiple outputs in Python API). I can access any layer by using the find_all_by_name method, but when I try to evaluate hidden layers using the eval method, it expects me to provide the layer's immediate input, rather than relevant network's inputs:

The BrainScript image reader supports horizontal flipping of training images. Is this also supported in Python? I did not see anything about it. Thx! https://github.com/Microsoft/CNTK/wiki/BrainScript-Image-reader

I am building an LSTM that handles several parallel sequences, and I'm struggling to find any brainscript example that handles dynamic axes.

In my specific case, an example consists of a binary label and N sequences, where each sequence i has a fixed length (but may differ for j<>i). For example, sequence 1 is always length 1024, sequence 2 is length 4096, sequence 3 is length 1024. I am expressing these sequences by packing them in parallel in the CNTK text format: