Improve fit_mesh tutorial

FluxML · Oct 7, 2020 · a81c652 · a81c652 · nirmal-suthar · Oct 8, 2020
1 parent 71a42d7
commit a81c652
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 3 deletions.
diff --git a/docs/src/tutorials/fit_mesh.md b/docs/src/tutorials/fit_mesh.md
@@ -99,6 +99,7 @@ end
 ```julia
 lr = 1.0
 opt = Flux.Optimise.Momentum(lr, 0.9)
+_offset = zeros(Float32, size(get_verts_packed(src))...)
 ```
 
 ## Using GPU for fast training [**Optional**]
@@ -108,7 +109,7 @@ function which is exactly the same syntax as Flux.
 ```julia
 tgt = tgt |> gpu
 src = src |> gpu
-_offset = zeros(Float32, size(get_verts_packed(src))...) |> gpu
+_offset = _offset |> gpu
 ```
 
 ## Optimizing the offset array
@@ -179,4 +180,3 @@ metrics.
 ---
 
 *This page was generated using [Literate.jl](https://github.com/fredrikekre/Literate.jl).*
-
diff --git a/examples/fit_mesh.jl b/examples/fit_mesh.jl
@@ -86,14 +86,15 @@ end
 
 lr = 1.0
 opt = Flux.Optimise.Momentum(lr, 0.9)
+_offset = zeros(Float32, size(get_verts_packed(src))...)
 
 # ## Using GPU for fast training [**Optional**]
 # We can convert the TriMesh structure to GPU or CPU using`gpu` and `cpu`
 # function which is exactly the same syntax as Flux.
 
 tgt = tgt |> gpu
 src = src |> gpu
-_offset = zeros(Float32, size(get_verts_packed(src))...) |> gpu
+_offset = _offset |> gpu
 
 # ## Optimizing the offset array
 # We first initialize offset array as zeros, hence deformed mesh is equivalent to