GitLab

Co-authored-by: moto <855818+mthrok@users.noreply.github.com>

* Switch to cmake for build
* Hide symbols

modified amplitude_to_DB to clamp per-item when a batch is provided
Co-authored-by: Vincent QB <vincentqb@users.noreply.github.com>

* Distinguish get_encodinginfo for Tensor I/O and save output

* Isolate get_tensor_encodinginfo so as not to use the same helper function

Co-authored-by: Prabhat Roy <prabhatroy@fb.com>

So that when a test fails in flaky manner, we can look into the difference.

Co-authored-by: Prabhat Roy <prabhatroy@fb.com>

Co-authored-by: Prabhat Roy <prabhatroy@fb.com>

[`sox_format_t.tell_off`](https://fossies.org/dox/sox-14.4.2/structsox__format__t.html#a2016a9fa839f3139e3c2f64381b0c445) should be representing current offset in file, but there are cases it does not. This was causing segmentation fault in some cases. This PR fixes it by replacing it with `ftell` call and add extra check so that if the same thing should happen, it will throw runtime error instead of segmentation fault.

fix 'file-does-not-end-in-newline' and typo.

Co-authored-by: krishnakalyan3 <skalyan@cloudera.com>

* Do not install sox system-wide, so that it will never get mixed up with static one torchaudio builds
* Do not install cmake system-wide
* Move libsndfile installation to Docker image

Changing the test Docker image for GPU tests to use Nvidia Docker `devel` image. This is necessary for #1187. 

TorchScript BC test checks that if dumped object from the previous version is compatible with the latest master. However, PyTorch does not guarantee backward compatibility of serialized TorchScript object across versions. Domain libraries do not guarantee this either.