v0.15.0 -Some API clean up; Added support for ONNX model conversion. #82
janpfeifer
announced in
Announcements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
graph:MatMul, with semantics similar tonumpy.matmulExpandDimstoInsertAxesand addedExpandAxes: the oldExpandDimshad a slightly differentsemantics than the usual (numpy)
expand_dimsthat I hadn't realized. The name change reflect that difference,and the new
ExpandAxesmatch the more common semantics ofexpand_dims. Added proper documentation.BREAKING CHANGE: easy to convert, but breaking anyway: it requires attention.
We defined a deprecated 'ExpandedDims' that maps to
InsertAxes, but it will be removed on the next release.node.ConstantValue,node.IsConstantExpression.context:ExecOnce: it was missing the variadic args for the computation graph.InspectVariableInScopeandInspectVariablerenamed toGetVariableandGetVariableByScopeAndNamerespectively. Alias to the older names left for compatibility (and marked as deprecated), but they
will be removed in future versions.
tensors:Tensor.Summary(precision int)to pretty-print a summary of the values of a tensor, numpy-like.See https://github.com/gomlx/onnx-gomlx for more on ONNX conversion to GoMLX.
This discussion was created from the release v0.15.0 -Some API clean up; Added support for ONNX model conversion..
Beta Was this translation helpful? Give feedback.
All reactions