hydrospheredata / hydro-root-cause Goto Github PK

Due to this bug algorithm tends to overestimate coverage and underestimate precision. Since the algorithm guarantees to generate high-precision anchors and each of for each anchor part precision contribution is underestimated, it generates too long and complex anchors.

I think from hydrosphere.io this is Apache 2.0, can you add a License file?

Generated URL:

1. http://localhost/status/rise/3ac2fc05-86de-4a1e-9478-0873a83af876
2. http://rootcause_http/fetch_result/rise/5d38449ff8a11febb432c117

Expected generated URL:

1. http://localhost/rootcause/status/rise/3ac2fc05-86de-4a1e-9478-0873a83af876
2. http://localhost/rootcause/fetch_result/rise/5d38449ff8a11febb432c117

RISE code is available only in notebook format. It needs to be restructured and rewritten into python module

If a user sends a tensor without specified TensorShapeProto, Root Cause service returns {state: Failed, description: 'tensor_shape'}.

e.g. consider creating tensor age_tensor = hs.TensorProto(dtype=hs.DT_INT64, int64_val=[32]) without specifying tensor_shape parameter. Sending this tensor to cluster is perfectly fine, monitoring and inference works. However, RootCause will return beforementioned error.

This issue can be fixed by adding a tenshor_shape parameter to the tensor constructor:
age_tensor = hs.TensorProto(dtype=hs.DT_INT64, int64_val=[32], tensor_shape=hs.TensorShapeProto())

So, the question is: can it be safely assumed that every tensor without tensor_shape param has hs.TensorShapeProto() shape?