Albert Zeyer

Submitted by pelya Since Albert added some progress indicator for map/mod loading which shows OLX not crashed, I think this one can be closed - please test.

Submitted by albertzeyer I moved that to feature requests and put the priority even lower. It's not really fixed and final yet as OLX itself is not really responding, the...

In RETURNN, we also have this capability. More specifically, we operate on sub-epochs. The user specifies the number of random partitions of a dataset. E.g. for Librispeech, we use 20,...

It looks like the channel number is totally off and also changes. This is strange. I thing the decoder channel number should not change. This indicates memory corruption, i.e. some...

What same error? The main error indicating that there is something wrong was the sample rate converter creation error. It couldn't be created because the number of channels was way...

To be fair, it is not recommended to write code like that, as you might get the case where `j` is not defined, if the loop never enters (e.g. iterating...

In LLDB where it hangs: ``` (lldb) thread backtrace all * thread #1, queue = 'com.apple.main-thread', stop reason = signal SIGSTOP * frame #0: 0x000000018382a270 libsystem_kernel.dylib`__psynch_cvwait + 8 frame #1:...

When executing on CPU, I get an actual exception: ``` TensorFlow exception: Graph execution error: indices[1] = 7 is not in [0, 7) [[{{node output/rec/enc_value_gather/GatherV2}}]] ``` And this is probably...

Thanks for creating the gist. However, as I said, it is crucial to run it on Apple M1 hardware to reproduce it. In your gist, I see that you use...

> **There is currently no official GPU support for MacOS.** It's better to post this issue [here](https://github.com/apple/tensorflow_macos/issues) for fast resolution. Thank you! It says: > This repository has been archived...