kurtamohler

@pytorchbot merge

I started an RFC (https://github.com/pytorch/rfcs/pull/43). So far I'm just focusing on listing out all the requirements the logging system needs. I'd appreciate any comments, especially about which requirements I haven't...

Here is a basic plan for pieces of the new logging system that can we can start implementing soon. Specifics about them can be found in the RFC (https://github.com/pytorch/rfcs/pull/43). Alban...

I don't get a failure in my environment either unfortunately. It would be useful to know exactly what all the args, kwargs, and result values are when it fails, for...

If `max_norm != None`, then `torch.nn.functional.embedding` calls `torch.embedding_renorm_`, which is implemented [here](https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/native/cuda/Embedding.cu#L347) for CUDA and [here](https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/native/Embedding.cpp#L153) for CPU, to normalize each row vector in the embedding matrix (the `weights` argument...

It looks like `opmath_t` is always set to `using opmath_t = at::opmath_type;`, and in `aten/src/ATen/OpMathType.h` we have: Click to expand ```cpp template struct OpMathType { using type = scalar_t; };...

Sorry, I forgot all about this. I will fix it up in the next couple days

I've added the warning and updated the test to check it

Given that we now have `TORCH_CHECK_TENSOR_ALL`, rather than a `TORCH_CHECK` that accepts a Tensor, I suppose I should plan to add `torch._check_tensor_all` in Python. Alternatively, we could easily make `torch._check`...

Where would be a good place to document the Python and C++ logging APIs? The simple answer is to just keep the documentation as comments in the code, but they're...