diff --git a/.buildkite/pipeline.yml b/.buildkite/pipeline.yml
index 5bb1dbfe3c..59538d1772 100644
--- a/.buildkite/pipeline.yml
+++ b/.buildkite/pipeline.yml
@@ -10,7 +10,6 @@ steps:
       queue: "juliagpu"
       cuda: "*"
     env:
-      JULIA_CUDA_USE_BINARYBUILDER: "true"
       FLUX_TEST_CUDA: "true"
       FLUX_TEST_CPU: "false"
     timeout_in_minutes: 60
diff --git a/Project.toml b/Project.toml
index f40a8cdab7..9c332a8fc4 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,6 +1,6 @@
 name = "Flux"
 uuid = "587475ba-b771-5e3f-ad9e-33799f191a9c"
-version = "0.14.8"
+version = "0.14.9"
 
 [deps]
 Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
diff --git a/ext/FluxCUDAExt/functor.jl b/ext/FluxCUDAExt/functor.jl
index e8a89c5553..dc8649fff0 100644
--- a/ext/FluxCUDAExt/functor.jl
+++ b/ext/FluxCUDAExt/functor.jl
@@ -29,6 +29,10 @@ adapt_storage(to::FluxCUDAAdaptor, x::AbstractRNG) =
 
 # TODO: figure out the correct design for OneElement
 adapt_storage(to::FluxCUDAAdaptor, x::Zygote.OneElement) = CUDA.cu(collect(x))
+# Patch for GPU support until we can make OneElement smarter
+if isdefined(Zygote.ChainRules, :OneElement)
+    adapt_storage(to::FluxCUDAAdaptor, x::Zygote.ChainRules.OneElement) = CUDA.cu(collect(x))
+end
 
 adapt_storage(to::FluxCPUAdaptor, x::T) where T <: CUDA.CUSPARSE.CUDA.CUSPARSE.AbstractCuSparseMatrix = adapt(Array, x)
 adapt_storage(to::FluxCPUAdaptor, x::CUDA.RNG) = Random.default_rng()