vargonis
vargonis
This issue regards branch rv_gpu. Consider the following: ```julia using Soss using CuArrays CuArrays.allowscalar(false) function gpu(trace) d = Dict(k => trace[k] for k in keys(trace)) for (k,v) in d v...
To reproduce: using IJulia on Atom via Hydrogen, evaluate `d = Dict(:x=>1, :y=>2)` and then type `d[:`. Autocompletion triggers and fails with the following error: ``` Dict{Symbol,Int64} with 2 entries:...
I'm getting inconsistent results among different calls for a depthwise convolution, with frequent `NaN`s. It can be reproduced, for instance, with: ``` using Flux c = DepthwiseConv((3,3), 3=>6, relu; pad=1)...
One has ```julia julia> typeof(rand(Uniform{Float32}(0f0,1f0))) Float64 ``` Similar inconsistencies occurs with `Exponential` and `Gamma`, although not with `Normal`. Relatedly: ```julia julia> eltype(Uniform{Float32}(0f0,1f0)) Float64 ``` This is rather easy to fix....
I tried to run the Cuda server from within a container, but a thread panics: ``` running /workspace/aici/target/release/rllm-cuda --verbose --aicirt /workspace/aici/target/release/aicirt -m microsoft/phi-2@d3186761bf5c4409f7679359284066c25ab668ee -t phi -w /workspace/aici/rllm/rllm-cuda/expected/phi-2/cats.safetensors --host 0.0.0.0 INFO...
The following code, which works well with Zygote, produces the detailed unexpected behavior: ``` using Flux.Tracker f(x) = [x^2+1, -x^3] y, back = Tracker.forward(f, 1) back([0,1]) # (-3.0 (tracked),) back(1)...
1. It is proposed that `find_element` and `get_elements_by_tagname` accept, besides an `XMLElement`, also an `XMLDocument` as first argument. 2. A method `find_element(x, ns::Vector{
``` A = rand(3,4) sum(I -> A[I], CartesianIndices(A); dims=1) # ERROR: BoundsError: attempt to access 3×4 Matrix{Float64} at index [0, 0] sum(I -> A[I], CartesianIndices(A); dims=1, init=0.) # now it...
It might be useful to have a mask that slightly generalizes `LengthMask`, in that it would not only specify an end index (the useful length of an input sequence in...