Wrap MPS(Graph) neural network primitives

[maleadt]

Closes #210
Interfaced with from FluxML/NNlib.jl#725

[github-actions]

Metal Benchmarks

Details

Benchmark suite	Current: 6e659a4	Previous: a9f57f4	Ratio
array/accumulate/Float32/1d	826917 ns	844374.5 ns	0.98
array/accumulate/Float32/dims=1	985833 ns	984417 ns	1.00
array/accumulate/Float32/dims=1L	8595750 ns	8591521 ns	1.00
array/accumulate/Float32/dims=2	1274041 ns	1271459 ns	1.00
array/accumulate/Float32/dims=2L	4020250.5 ns	4014000 ns	1.00
array/accumulate/Int64/1d	906750 ns	884917 ns	1.02
array/accumulate/Int64/dims=1	1133250 ns	1130979.5 ns	1.00
array/accumulate/Int64/dims=1L	9578396 ns	9575500 ns	1.00
array/accumulate/Int64/dims=2	1457000 ns	1454666 ns	1.00
array/accumulate/Int64/dims=2L	6591875.5 ns	6576584 ns	1.00
array/broadcast	368125 ns	363500 ns	1.01
array/construct	3292 ns	3375 ns	0.98
array/permutedims/2d	631666 ns	598833 ns	1.05
array/permutedims/3d	1088854.5 ns	1082917 ns	1.01
array/permutedims/4d	1808354 ns	1640708 ns	1.10
array/private/copy	418791 ns	412812.5 ns	1.01
array/private/copyto!/cpu_to_gpu	384291 ns	339959 ns	1.13
array/private/copyto!/gpu_to_cpu	382666 ns	381834 ns	1.00
array/private/copyto!/gpu_to_gpu	357000 ns	356709 ns	1.00
array/private/iteration/findall/bool	1117187.5 ns	1126666 ns	0.99
array/private/iteration/findall/int	1280416.5 ns	1275208 ns	1.00
array/private/iteration/findfirst/bool	1412042 ns	1396875 ns	1.01
array/private/iteration/findfirst/int	1442042 ns	1429354 ns	1.01
array/private/iteration/findmin/1d	1575333 ns	1570583.5 ns	1.00
array/private/iteration/findmin/2d	1268125 ns	1253666 ns	1.01
array/private/iteration/logical	1829083 ns	1816000 ns	1.01
array/private/iteration/scalar	2793083 ns	2845542 ns	0.98
array/random/rand/Float32	615959 ns	610125 ns	1.01
array/random/rand/Int64	631042 ns	629209 ns	1.00
array/random/rand!/Float32	530541 ns	524583 ns	1.01
array/random/rand!/Int64	494667 ns	493959 ns	1.00
array/random/randn/Float32	575625 ns	568291 ns	1.01
array/random/randn!/Float32	479083.5 ns	477917 ns	1.00
array/reductions/mapreduce/Float32/1d	702833 ns	703625 ns	1.00
array/reductions/mapreduce/Float32/dims=1	482625 ns	481209 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	687667 ns	686666 ns	1.00
array/reductions/mapreduce/Float32/dims=2	494875 ns	491000 ns	1.01
array/reductions/mapreduce/Float32/dims=2L	1285645.5 ns	1281250 ns	1.00
array/reductions/mapreduce/Int64/1d	896875 ns	892979 ns	1.00
array/reductions/mapreduce/Int64/dims=1	772937.5 ns	773083 ns	1.00
array/reductions/mapreduce/Int64/dims=1L	1083541 ns	1085417 ns	1.00
array/reductions/mapreduce/Int64/dims=2	932667 ns	925083 ns	1.01
array/reductions/mapreduce/Int64/dims=2L	2177020.5 ns	2169417 ns	1.00
array/reductions/reduce/Float32/1d	686459 ns	698125 ns	0.98
array/reductions/reduce/Float32/dims=1	480583 ns	480167 ns	1.00
array/reductions/reduce/Float32/dims=1L	686145.5 ns	686166 ns	1.00
array/reductions/reduce/Float32/dims=2	495042 ns	488625 ns	1.01
array/reductions/reduce/Float32/dims=2L	1284791 ns	1275417 ns	1.01
array/reductions/reduce/Int64/1d	894583 ns	894541.5 ns	1.00
array/reductions/reduce/Int64/dims=1	773625 ns	772250 ns	1.00
array/reductions/reduce/Int64/dims=1L	1078895.5 ns	1078708 ns	1.00
array/reductions/reduce/Int64/dims=2	932500 ns	927375 ns	1.01
array/reductions/reduce/Int64/dims=2L	2156084 ns	2149500 ns	1.00
array/shared/copy	152750 ns	153083 ns	1.00
array/shared/copyto!/cpu_to_gpu	39916 ns	40250 ns	0.99
array/shared/copyto!/gpu_to_cpu	40000 ns	39500 ns	1.01
array/shared/copyto!/gpu_to_gpu	40583 ns	40792 ns	0.99
array/shared/iteration/findall/bool	1121083 ns	1117167 ns	1.00
array/shared/iteration/findall/int	1285166 ns	1278417 ns	1.01
array/shared/iteration/findfirst/bool	1119542 ns	1114417 ns	1.00
array/shared/iteration/findfirst/int	1154417 ns	1142479.5 ns	1.01
array/shared/iteration/findmin/1d	1305042 ns	1294500 ns	1.01
array/shared/iteration/findmin/2d	1273500 ns	1276333 ns	1.00
array/shared/iteration/logical	1647000 ns	1640958 ns	1.00
array/shared/iteration/scalar	3526 ns	3453.125 ns	1.02
integration/byval/reference	1139833 ns	1144334 ns	1.00
integration/byval/slices=1	1134500 ns	1136833 ns	1.00
integration/byval/slices=2	2048895.5 ns	2052958 ns	1.00
integration/byval/slices=3	7152417 ns	7070625 ns	1.01
integration/metaldevrt	441125 ns	437125 ns	1.01
kernel/indexing	357333 ns	356542 ns	1.00
kernel/indexing_checked	531208 ns	530458 ns	1.00
kernel/launch	1954.2 ns	1966.7 ns	0.99
kernel/rand	529625 ns	519917 ns	1.02
latency/import	1852920959 ns	1855549750 ns	1.00
latency/precompile	36991825854 ns	36949704062.5 ns	1.00
latency/ttfp	2069514833 ns	2075329333 ns	1.00
metal/synchronization/context	535.15625 ns	519.4300518134716 ns	1.03
metal/synchronization/stream	275.7936507936508 ns	275.4182389937107 ns	1.00

This comment was automatically generated by workflow using github-action-benchmark.

[codecov]

Codecov Report

❌ Patch coverage is 99.29577% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 85.75%. Comparing base (eed9d53) to head (6e659a4).
⚠️ Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/mpsgraphs/nn.jl	99.29%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #854      +/-   ##
==========================================
+ Coverage   84.38%   85.75%   +1.36%     
==========================================
  Files          70       71       +1     
  Lines        4196     4478     +282     
==========================================
+ Hits         3541     3840     +299     
+ Misses        655      638      -17     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[christiangnrd]

Supersedes #745?