RFC #112
RFC #112
Conversation
RFC.md
Outdated
| w = z.sum() | ||
| ``` | ||
|
|
||
| Here we already see a couple differences between NumPy and PyTorch. The most |
There was a problem hiding this comment.
I'd suggest to move a summary of differences to a section nearer to the end of the document, and forward-reference it here. It reads better if we focus on the main use case and 95% where behavior is matching.
There was a problem hiding this comment.
I pushed some changes. I think that leaving these in this example and moving the FKA ## DTypes section right after this section makes these points here have more meaning.
This is how I had it at the beginning, which makes more sense, but then I moved the dtypes section down the document and left this part a bit on its own.
RFC.md
Outdated
| keyword-only argument. It is for this reason that, in PrimTorch, we were able | ||
| to implement it as | ||
| [a decorator](https://github.com/pytorch/pytorch/blob/ce4df4cc596aa10534ac6d54912f960238264dfd/torch/_prims_common/wrappers.py#L187-L282). | ||
| This is not the case in NumPy. In NumPy `out` is a positional arg that is often |
There was a problem hiding this comment.
That's going on my list of things to fix for NumPy 2.0:)
There was a problem hiding this comment.
Not sure what I meant two weeks ago, but this isn't true:
>>> x = np.ones(3)
>>> np.sin(x, x)
array([0.84147098, 0.84147098, 0.84147098])
>>> x = np.ones(3)
>>> np.sin(x, out=x)
array([0.84147098, 0.84147098, 0.84147098])I'll change this to "can be used both as a positional and a keyword argument".
There was a problem hiding this comment.
yep, I meant "In NumPy, out can be specified as a positional arg"
RFC.md
Outdated
|
|
||
|
|
||
| The this project has a main goal as per the | ||
| [initial design document](https://docs.google.com/document/d/1gdUDgZNbumFORRcUaZUVw790CtNYweAM20C1fbWMNd8): |
There was a problem hiding this comment.
I'll try to fix that by upstreaming the relevant content here. Best for this RFC to be standalone.
|
Addressed the last review. Once @rgommers does a last pass, we'll publish this one in the RFC repo. |
I left a number of comments in bold throughout the text that occurred to me as I was writing this.