Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ConvTranspose errors with symmetric non-constant pad #2424

Open
piever opened this issue Apr 17, 2024 · 0 comments
Open

ConvTranspose errors with symmetric non-constant pad #2424

piever opened this issue Apr 17, 2024 · 0 comments

Comments

@piever
Copy link
Contributor

piever commented Apr 17, 2024

Description

The ConvTranspose layer errors if given a symmetric non-constant pad (that is, length(pad) == ndims(w) - 2), whereas pad::Int (constant pad) and length(pad) == 2 * (ndims(w) - 2) (non-symmetric non-constant pad) work fine.

MWE:

using Flux

c = ConvTranspose((3, 3), 1 => 1, pad = (1, 0))

x = randn(Float32, 5, 5, 1, 16)

c(x)

errors with

ERROR: MethodError: no method matching DenseConvDims(::Tuple{Int64, Int64, Int64}, ::NTuple{4, Int64}; stride::Tuple{Int64, Int64}, padding::Tuple{Int64, Int64}, dilation::Tuple{Int64, Int64}, groups::Int64)

Closest candidates are:
  DenseConvDims(::Tuple{Vararg{Int64, N}}, ::Tuple{Vararg{Int64, K}}, ::Int64, ::Int64, ::Int64, ::Tuple{Vararg{Int64, S}}, ::Tuple{Vararg{Int64, P}}, ::Tuple{Vararg{Int64, D}}, ::Bool) where {N, K, S, P, D} got unsupported keyword arguments "stride", "padding", "dilation", "groups"
   @ NNlib C:\Users\pietro\.julia\packages\NNlib\O0zGY\src\dim_helpers\DenseConvDims.jl:7
  DenseConvDims(::Tuple{Vararg{T, M}} where T, ::Tuple{Vararg{T, M}} where T; stride, padding, dilation, groups, flipkernel) where M
   @ NNlib C:\Users\pietro\.julia\packages\NNlib\O0zGY\src\dim_helpers\DenseConvDims.jl:20

Stacktrace:
 [1] conv_transpose_dims(c::ConvTranspose{2, 2, typeof(identity), Array{Float32, 4}, Vector{Float32}}, x::Array{Float32, 4})
   @ Flux C:\Users\pietro\.julia\packages\Flux\Wz6D4\src\layers\conv.jl:323
 [2] (::ConvTranspose{2, 2, typeof(identity), Array{Float32, 4}, Vector{Float32}})(x::Array{Float32, 4})
   @ Flux C:\Users\pietro\.julia\packages\Flux\Wz6D4\src\layers\conv.jl:336
 [3] top-level scope
   @ REPL[6]:1

Environment

The above reproducer works in Julia 1.10.2 on a fresh environment

(jl_tMb72B) pkg> st
Status `C:\Users\pietro\AppData\Local\Temp\jl_tMb72B\Project.toml`
  [587475ba] Flux v0.14.15

Suggestion

I suspect the problem is in combined_pad (here and here). It assumes that c.pad has length 2 * (ndims(c.weight) - 2), which is not the case.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants