denizyuret

Continuing from: https://groups.google.com/d/topic/julia-dev/NqPz4f_0VLg/discussion I get this error intermittently on exit from Julia. I was able to trace it to the finalizer of CudaPtr (pointer.jl:44), in particular to the statement haskey(cuda_ptrs,...

Types with CudaArray elements cannot be saved to JLD because of pointer exception

3

I'd like to fix this so people can save and load machine learning models that use CudaArray's without having to explicitly copy everything to cpu. Is the right way to...

How about a Knet org?

5

My students have been asking for a Knet github organization for a while to contribute projects that are large to be just examples. I did not think there was enough...

``` julia> tsne(randn(Float64,500,100)); Computing t-SNE100%|████████████████████████████████████| Time: 0:00:00 KL_divergence: 2.2364 julia> tsne(randn(Float32,500,100)); Computing t-SNE100%|████████████████████████████████████| Time: 0:01:23 KL_divergence: 2.2551 ```

importall BLAS for generic code

5

Hi Nick, Currently CUBLAS does not export or modify any existing functions, but provides separate CUBLAS versions like CUBLAS.axpy! etc. (If I understand things correctly). If you add the line:...

transpose not working

1

Thanks for porting this library! I had started doing this a few months ago and gave up after I discovered a problem with transpose. I am not sure if NVIDIA...

This is a replacement for https://github.com/EleutherAI/lm-evaluation-harness/pull/1039 reimplementing the no-softmax feature to support globally normalized models discussed in https://github.com/EleutherAI/lm-evaluation-harness/issues/960 using new entries in `MODEL_REGISTRY` instead of a command line option as...

globally normalized models

1

Globally normalized models score sequences using the sum of unnormalized logits, as opposed to locally normalized ones which take a log_softmax at each token position before summing to compute the...

Cannot create custom image in Julia 1.10: Method overwriting is not permitted during Module precompilation

6

Following the instructions at https://pyjulia.readthedocs.io/en/latest/sysimage.html with Julia 1.10.0 I get the error given in the subject. Here is the full output: https://gist.github.com/denizyuret/dcf47516b0379ce1f7fce8069610d039. The instructions seem to work with Julia 1.9.4....