Update to CUTENSOR 1.6 and cuQuanty 22.07
@kshyatt could you finish this? I only wanted to update CUTENSOR, but cuQuantum throws a fit then (version mismatch errors) so needs to be updated too. There's been many changes between the 0.1 versions you created wrappers for, and the 1.0 versions of CUSTATEVEC and CUTENSORNET. I already fixed CUSTATEVEC, but CUTENSORNET fails with an API error:
[2022-10-11 13:14:52][cuTENSORNet][57010][Error][cutensornetContractionOptimize] No valid path could be found with specified constraints.
tuning = NoAutoTune(): Error During Test at /home/tim/Julia/pkg/CUDA/lib/cutensornet/test/runtests.jl:30
Got exception outside of a @test
CUTENSORNETError: all hyper samples failed for one or more errors (enable LOGs via export CUTENSORNET_LOG_LEVEL= > 1 for details). (code 24, CUTENSORNET_STATUS_ALL_HYPER_SAMPLES_FAILED)
Stacktrace:
[1] throw_api_error(res::CUTENSORNET.cutensornetStatus_t)
@ CUTENSORNET ~/Julia/pkg/CUDA/lib/cutensornet/src/error.jl:61
[2] macro expansion
@ ~/Julia/pkg/CUDA/lib/cutensornet/src/error.jl:74 [inlined]
[3] cutensornetContractionOptimize(handle::Ptr{Nothing}, descNet::CUTENSORNET.CuTensorNetworkDescriptor, optimizerConfig::CUTENSORNET.CuTensorNetworkContractionOptimizerConfig, workspaceSizeConstraint::Int64, optimizerInfo::CUTENSORNET.CuTensorNetworkContractionOptimizerInfo)
@ CUTENSORNET ~/Julia/pkg/CUDA/lib/utils/call.jl:26
[4] rehearse_contraction(tn::CuTensorNetwork{Float32}, max_workspace_size::Int64, optimizer_conf::CUTENSORNET.OptimizerConfig)
@ CUTENSORNET ~/Julia/pkg/CUDA/lib/cutensornet/src/tensornet.jl:7
[5] rehearse_contraction(tn::CuTensorNetwork{Float32}, max_workspace_size::Int64)
@ CUTENSORNET ~/Julia/pkg/CUDA/lib/cutensornet/src/tensornet.jl:3
Release notes here: https://docs.nvidia.com/cuda/cuquantum/cutensornet/release_notes.html#cutensornet-v1-1-1. One change that still needs to happen is the workspace changes, https://docs.nvidia.com/cuda/cuquantum/cutensornet/getting_started.html#create-workspace-descriptor-and-allocate-workspace-memory; I already updated the wrappers (so it should error), but that code only gets invoked after the above API error, so something else is up.
Yeah sure I can take over!
Codecov Report
Base: 74.39% // Head: 70.59% // Decreases project coverage by -3.79% :warning:
Coverage data is based on head (
94c3f1a) compared to base (ebba936). Patch coverage: 0.00% of modified lines in pull request are covered.
Additional details and impacted files
@@ Coverage Diff @@
## master #1623 +/- ##
==========================================
- Coverage 74.39% 70.59% -3.80%
==========================================
Files 131 131
Lines 10141 9665 -476
==========================================
- Hits 7544 6823 -721
- Misses 2597 2842 +245
| Impacted Files | Coverage Δ | |
|---|---|---|
| lib/custatevec/src/CUSTATEVEC.jl | 0.00% <ø> (ø) |
|
| lib/custatevec/src/error.jl | 0.00% <0.00%> (ø) |
|
| lib/custatevec/src/libcustatevec.jl | 0.00% <0.00%> (ø) |
|
| lib/custatevec/src/statevec.jl | 0.00% <0.00%> (ø) |
|
| lib/custatevec/src/types.jl | 0.00% <0.00%> (ø) |
|
| lib/cutensornet/src/CUTENSORNET.jl | 0.00% <ø> (ø) |
|
| lib/cutensornet/src/error.jl | 0.00% <0.00%> (ø) |
|
| lib/cutensornet/src/libcutensornet.jl | 0.00% <0.00%> (ø) |
|
| lib/cutensornet/src/tensornet.jl | 0.00% <ø> (ø) |
|
| lib/cutensornet/src/types.jl | 0.00% <0.00%> (ø) |
|
| ... and 84 more |
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.
:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
Thanks! Note that https://github.com/JuliaGPU/CUDA.jl/pull/1624 changed where files are, so this will need a rebase.
Superseded by #1636 #1638 #1639.