I am getting the following error
(CompileError) deps/axon/lib/axon/loop.ex:469: the do-block in while must return tensors with the same shape, type, and names as the initial arguments.
While trying to enforce the following policy
policy = Axon.MixedPrecision.create_policy( params: {:f, 64}, compute: {:f, 64}, output: {:f, 64} )
The error seems to be due to the gradient_state below initiated as :f32
|
gradient_state: zeros_like(model_state, type: :f32), |
I am getting the following error
(CompileError) deps/axon/lib/axon/loop.ex:469: the do-block in while must return tensors with the same shape, type, and names as the initial arguments.While trying to enforce the following policy
policy = Axon.MixedPrecision.create_policy( params: {:f, 64}, compute: {:f, 64}, output: {:f, 64} )The error seems to be due to the
gradient_statebelow initiated as:f32axon/lib/axon/loop.ex
Line 360 in ddc49cc