update metric computations (#16)

Author: Dando18

analysis/all-metrics.sh

@@ -8,7 +8,7 @@ python metrics.py ../results/a8724ee8/starcoderbase_prompted_temp0.2/results.csv
 python metrics.py ../results/a8724ee8/phind-v2_prompted_temp0.2/results.csv --model-name Phind-V2 --output ../results/a8724ee8/phind-v2_prompted_temp0.2/metrics.csv
 python metrics.py ../results/a8724ee8/gpt-3.5_temp0.2/results.csv --model-name GPT-3.5 --output ../results/a8724ee8/gpt-3.5_temp0.2/metrics.csv
 python metrics.py ../results/a8724ee8/gpt-4_temp0.2/results.csv --model-name GPT-4 --output ../results/a8724ee8/gpt-4_temp0.2/metrics.csv
-
+(cd ../results/a8724ee8 && head -n 1 codellama-7b-hf_prompted_temp0.2/metrics.csv > all-metrics.csv && tail -n+2 -q */metrics.csv >> all-metrics.csv)
 
 # mpi scaling metrics
 python metrics-scaling.py ../results/a8724ee8/codellama-7b-hf_prompted_temp0.2/results.csv --model-name CodeLlama-7B -k 1 -n 1 2 4 8 16 32 64 128 256 512 --execution-model mpi --output ../results/a8724ee8/codellama-7b-hf_prompted_temp0.2/metrics-scaling-mpi.csv
@@ -18,17 +18,17 @@ python metrics-scaling.py ../results/a8724ee8/starcoderbase_prompted_temp0.2/res
 python metrics-scaling.py ../results/a8724ee8/phind-v2_prompted_temp0.2/results.csv --model-name Phind-V2 -k 1 -n 1 2 4 8 16 32 64 128 256 512 --execution-model mpi --output ../results/a8724ee8/phind-v2_prompted_temp0.2/metrics-scaling-mpi.csv
 python metrics-scaling.py ../results/a8724ee8/gpt-3.5_temp0.2/results.csv --model-name GPT-3.5 -k 1 -n 1 2 4 8 16 32 64 128 256 512 --execution-model mpi --output ../results/a8724ee8/gpt-3.5_temp0.2/metrics-scaling-mpi.csv
 python metrics-scaling.py ../results/a8724ee8/gpt-4_temp0.2/results.csv --model-name GPT-4 -k 1 -n 1 2 4 8 16 32 64 128 256 512 --execution-model mpi --output ../results/a8724ee8/gpt-4_temp0.2/metrics-scaling-mpi.csv
-
+(cd ../results/a8724ee8 && head -n 1 codellama-7b-hf_prompted_temp0.2/metrics-scaling-mpi.csv > all-metrics-scaling-mpi.csv && tail -n+2 -q */metrics-scaling-mpi.csv >> all-metrics-scaling-mpi.csv)
 
 # omp scaling metrics
-python metrics-scaling.py ../results/a8724ee8/codellama-7b-hf_prompted_temp0.2/results.csv --model-name CodeLlama-7B -k 1 -n 1 2 4 8 16 32 64 --execution-model omp --output ../results/a8724ee8/codellama-7b-hf_prompted_temp0.2/metrics-scaling-omp.csv
-python metrics-scaling.py ../results/a8724ee8/codellama-13b-hf_prompted_temp0.2/results.csv --model-name CodeLlama-13B -k 1 -n 1 2 4 8 16 32 64 --execution-model omp --output ../results/a8724ee8/codellama-13b-hf_prompted_temp0.2/metrics-scaling-omp.csv
-python metrics-scaling.py ../results/a8724ee8/codellama-34b-hf_prompted_temp0.2/results.csv --model-name CodeLlama-34B -k 1 -n 1 2 4 8 16 32 64 --execution-model omp --output ../results/a8724ee8/codellama-34b-hf_prompted_temp0.2/metrics-scaling-omp.csv
-python metrics-scaling.py ../results/a8724ee8/starcoderbase_prompted_temp0.2/results.csv --model-name StarCoderBase -k 1 -n 1 2 4 8 16 32 64 --execution-model omp --output ../results/a8724ee8/starcoderbase_prompted_temp0.2/metrics-scaling-omp.csv
-python metrics-scaling.py ../results/a8724ee8/phind-v2_prompted_temp0.2/results.csv --model-name Phind-V2 -k 1 -n 1 2 4 8 16 32 64 --execution-model omp --output ../results/a8724ee8/phind-v2_prompted_temp0.2/metrics-scaling-omp.csv
-python metrics-scaling.py ../results/a8724ee8/gpt-3.5_temp0.2/results.csv --model-name GPT-3.5 -k 1 -n 1 2 4 8 16 32 64 --execution-model omp --output ../results/a8724ee8/gpt-3.5_temp0.2/metrics-scaling-omp.csv
-python metrics-scaling.py ../results/a8724ee8/gpt-4_temp0.2/results.csv --model-name GPT-4 -k 1 -n 1 2 4 8 16 32 64 --execution-model omp --output ../results/a8724ee8/gpt-4_temp0.2/metrics-scaling-omp.csv
-
+python metrics-scaling.py ../results/a8724ee8/codellama-7b-hf_prompted_temp0.2/results.csv --model-name CodeLlama-7B -k 1 -n 1 2 4 8 16 32 --execution-model omp --output ../results/a8724ee8/codellama-7b-hf_prompted_temp0.2/metrics-scaling-omp.csv
+python metrics-scaling.py ../results/a8724ee8/codellama-13b-hf_prompted_temp0.2/results.csv --model-name CodeLlama-13B -k 1 -n 1 2 4 8 16 32 --execution-model omp --output ../results/a8724ee8/codellama-13b-hf_prompted_temp0.2/metrics-scaling-omp.csv
+python metrics-scaling.py ../results/a8724ee8/codellama-34b-hf_prompted_temp0.2/results.csv --model-name CodeLlama-34B -k 1 -n 1 2 4 8 16 32 --execution-model omp --output ../results/a8724ee8/codellama-34b-hf_prompted_temp0.2/metrics-scaling-omp.csv
+python metrics-scaling.py ../results/a8724ee8/starcoderbase_prompted_temp0.2/results.csv --model-name StarCoderBase -k 1 -n 1 2 4 8 16 32 --execution-model omp --output ../results/a8724ee8/starcoderbase_prompted_temp0.2/metrics-scaling-omp.csv
+python metrics-scaling.py ../results/a8724ee8/phind-v2_prompted_temp0.2/results.csv --model-name Phind-V2 -k 1 -n 1 2 4 8 16 32 --execution-model omp --output ../results/a8724ee8/phind-v2_prompted_temp0.2/metrics-scaling-omp.csv
+python metrics-scaling.py ../results/a8724ee8/gpt-3.5_temp0.2/results.csv --model-name GPT-3.5 -k 1 -n 1 2 4 8 16 32 --execution-model omp --output ../results/a8724ee8/gpt-3.5_temp0.2/metrics-scaling-omp.csv
+python metrics-scaling.py ../results/a8724ee8/gpt-4_temp0.2/results.csv --model-name GPT-4 -k 1 -n 1 2 4 8 16 32 --execution-model omp --output ../results/a8724ee8/gpt-4_temp0.2/metrics-scaling-omp.csv
+(cd ../results/a8724ee8 && head -n 1 codellama-7b-hf_prompted_temp0.2/metrics-scaling-omp.csv > all-metrics-scaling-omp.csv && tail -n+2 -q */metrics-scaling-omp.csv >> all-metrics-scaling-omp.csv)
 
 # kokkos scaling metrics
 python metrics-scaling.py ../results/a8724ee8/codellama-7b-hf_prompted_temp0.2/results.csv --model-name CodeLlama-7B -k 1 -n 1 2 4 8 16 32 --execution-model kokkos --output ../results/a8724ee8/codellama-7b-hf_prompted_temp0.2/metrics-scaling-kokkos.csv
@@ -37,4 +37,5 @@ python metrics-scaling.py ../results/a8724ee8/codellama-34b-hf_prompted_temp0.2/
 python metrics-scaling.py ../results/a8724ee8/starcoderbase_prompted_temp0.2/results.csv --model-name StarCoderBase -k 1 -n 1 2 4 8 16 32 --execution-model kokkos --output ../results/a8724ee8/starcoderbase_prompted_temp0.2/metrics-scaling-kokkos.csv
 python metrics-scaling.py ../results/a8724ee8/phind-v2_prompted_temp0.2/results.csv --model-name Phind-V2 -k 1 -n 1 2 4 8 16 32 --execution-model kokkos --output ../results/a8724ee8/phind-v2_prompted_temp0.2/metrics-scaling-kokkos.csv
 python metrics-scaling.py ../results/a8724ee8/gpt-3.5_temp0.2/results.csv --model-name GPT-3.5 -k 1 -n 1 2 4 8 16 32 --execution-model kokkos --output ../results/a8724ee8/gpt-3.5_temp0.2/metrics-scaling-kokkos.csv
-python metrics-scaling.py ../results/a8724ee8/gpt-4_temp0.2/results.csv --model-name GPT-4 -k 1 -n 1 2 4 8 16 32 --execution-model kokkos --output ../results/a8724ee8/gpt-4_temp0.2/metrics-scaling-kokkos.csv
+python metrics-scaling.py ../results/a8724ee8/gpt-4_temp0.2/results.csv --model-name GPT-4 -k 1 -n 1 2 4 8 16 32 --execution-model kokkos --output ../results/a8724ee8/gpt-4_temp0.2/metrics-scaling-kokkos.csv
+(cd ../results/a8724ee8 && head -n 1 codellama-7b-hf_prompted_temp0.2/metrics-scaling-kokkos.csv > all-metrics-scaling-kokkos.csv && tail -n+2 -q */metrics-scaling-kokkos.csv >> all-metrics-scaling-kokkos.csv)

analysis/metrics.py

@@ -98,7 +98,7 @@ def speedupk(df: pd.DataFrame, k: int, n: int) -> pd.DataFrame:
             (df["parallelism_model"] == "cuda") |
             (df["parallelism_model"] == "hip") |
             ((df["parallelism_model"] == "kokkos") & (df["num_threads"] == 32)) |
-            ((df["parallelism_model"] == "omp") & (df["num_threads"] == 64)) |
+            ((df["parallelism_model"] == "omp") & (df["num_threads"] == 32)) |
             ((df["parallelism_model"] == "mpi") & (df["num_procs"] == 512)) |
             ((df["parallelism_model"] == "mpi+omp") & (df["num_procs"] == 4) & (df["num_threads"] == 64))]
     df = df.copy()
@@ -153,17 +153,17 @@ def efficiencyk(df: pd.DataFrame, k: int, n: int) -> pd.DataFrame:
             (df["parallelism_model"] == "cuda") |
             (df["parallelism_model"] == "hip") |
             ((df["parallelism_model"] == "kokkos") & (df["num_threads"] == 32)) |
-            ((df["parallelism_model"] == "omp") & (df["num_threads"] == 64)) |
+            ((df["parallelism_model"] == "omp") & (df["num_threads"] == 32)) |
             ((df["parallelism_model"] == "mpi") & (df["num_procs"] == 512)) |
             ((df["parallelism_model"] == "mpi+omp") & (df["num_procs"] == 4) & (df["num_threads"] == 64))]
 
-    # set n_resources column to 1 for serial; 32 for kokkos; 64 for omp; 512 for mpi; 4*64 for mpi+omp;
+    # set n_resources column to 1 for serial; 32 for kokkos; 32 for omp; 512 for mpi; 4*64 for mpi+omp;
     # set it to problem_size for cuda and hip
     df["n_resources"] = 1
     df.loc[df["parallelism_model"] == "cuda", "n_resources"] = df["problem_size"]
     df.loc[df["parallelism_model"] == "hip", "n_resources"] = df["problem_size"]
     df.loc[df["parallelism_model"] == "kokkos", "n_resources"] = 32
-    df.loc[df["parallelism_model"] == "omp", "n_resources"] = 64
+    df.loc[df["parallelism_model"] == "omp", "n_resources"] = 32
     df.loc[df["parallelism_model"] == "mpi", "n_resources"] = 512
     df.loc[df["parallelism_model"] == "mpi+omp", "n_resources"] = 4*64
 

results/a8724ee8/tmp.csv renamed to results/a8724ee8/all-metrics-scaling-kokkos.csv

@@ -6,22 +6,19 @@ CodeLlama-13B,kokkos,histogram,0.49254226844721943,0.49254226844721943,0.4069104
 CodeLlama-13B,kokkos,reduce,1.1526706016062902,1.1526706016062902,2.131513786453558,1.065756893226779,4.0834481710452755,1.0208620427613189,7.970097397338397,0.9962621746672996,14.76651661769161,0.9229072886057256,21.760548045579046,0.6800171264243452
 CodeLlama-13B,kokkos,stencil,0.43833601423162,0.43833601423162,0.8767905895853975,0.43839529479269873,1.7612435990246864,0.4403108997561716,3.4767587943453737,0.4345948492931717,6.9669270153104135,0.43543293845690084,14.2522656251154,0.4453833007848563
 CodeLlama-13B,kokkos,transform,1.3923379061404613,1.3923379061404613,2.772874273538197,1.3864371367690984,5.535232842581403,1.3838082106453506,10.900981562175152,1.362622695271894,22.22756623699908,1.3892228898124426,44.52229817148291,1.391321817858841
-model,execution model,problem type,speedup_1@1,efficiency_1@1,speedup_2@1,efficiency_2@1,speedup_4@1,efficiency_4@1,speedup_8@1,efficiency_8@1,speedup_16@1,efficiency_16@1,speedup_32@1,efficiency_32@1
 CodeLlama-34B,kokkos,dense_la,0.46104268672267223,0.46104268672267223,0.8970556873070841,0.44852784365354204,1.8173239561862307,0.45433098904655767,3.4848213979602765,0.43560267474503456,7.0911014278008455,0.44319383923755284,13.074452561256287,0.40857664253925896
 CodeLlama-34B,kokkos,geometry,0.2255136175593122,0.2255136175593122,0.258016361499565,0.1290081807497825,0.389962216918869,0.09749055422971725,0.6823356393674852,0.08529195492093565,1.301496165371815,0.08134351033573843,2.5034951305118875,0.07823422282849649
 CodeLlama-34B,kokkos,graph,0.2813588647867369,0.2813588647867369,0.562697665917093,0.2813488329585465,1.1220333084177718,0.28050832710444296,2.2509511356418304,0.2813688919552288,4.451019674369042,0.2781887296480651,8.704165378354245,0.27200516807357017
 CodeLlama-34B,kokkos,reduce,0.8256506642395032,0.8256506642395032,1.2221525646976001,0.6110762823488001,2.0267125374848325,0.5066781343712081,3.6332410243356055,0.4541551280419507,6.981493950476832,0.436343371904802,13.449149742085751,0.42028592944017973
 CodeLlama-34B,kokkos,scan,1.0910775021560708,1.0910775021560708,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0
 CodeLlama-34B,kokkos,search,0.2152754313882857,0.2152754313882857,0.4309696112665852,0.2154848056332926,0.8622722595132744,0.2155680648783186,1.661552988598795,0.20769412357484937,2.505055087039688,0.1565659429399805,2.753305107567982,0.08604078461149944
 CodeLlama-34B,kokkos,transform,0.8652845230691931,0.8652845230691931,1.722771113356363,0.8613855566781815,3.438078236459027,0.8595195591147567,6.859029761231295,0.8573787201539119,13.896111264469466,0.8685069540293416,27.56260657775942,0.8613314555549819
-model,execution model,problem type,speedup_1@1,efficiency_1@1,speedup_2@1,efficiency_2@1,speedup_4@1,efficiency_4@1,speedup_8@1,efficiency_8@1,speedup_16@1,efficiency_16@1,speedup_32@1,efficiency_32@1
 CodeLlama-7B,kokkos,geometry,0.3989260416446082,0.3989260416446082,0.38341688036324106,0.19170844018162053,0.398769483501477,0.09969237087536925,0.3853735007711615,0.04817168759639519,0.40600981953501886,0.02537561372093868,0.38413071249338177,0.01200408476541818
 CodeLlama-7B,kokkos,graph,0.37573765800692493,0.37573765800692493,0.37584688697971325,0.18792344348985662,0.37600728799120847,0.09400182199780212,0.37572398134541374,0.04696549766817672,0.37573752902977553,0.02348359556436097,0.37553268545617535,0.01173539642050548
 CodeLlama-7B,kokkos,histogram,0.5697468782224298,0.5697468782224298,0.6979518632258975,0.34897593161294876,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0
 CodeLlama-7B,kokkos,reduce,0.8980571081545551,0.8980571081545551,1.2669323426313615,0.6334661713156807,1.9977467569992455,0.4994366892498114,3.4517109635203154,0.4314638704400394,6.320630111338264,0.3950393819586415,11.939398786482709,0.37310621207758465
 CodeLlama-7B,kokkos,stencil,0.41698213454476646,0.41698213454476646,0.7032890069930849,0.35164450349654247,1.5304014871427836,0.3826003717856959,3.047667874811251,0.3809584843514064,6.065632358294693,0.37910202239341834,11.970177673603153,0.37406805230009854
 CodeLlama-7B,kokkos,transform,1.2535370832205182,1.2535370832205182,2.468054586909754,1.234027293454877,4.878616879394713,1.2196542198486782,9.722920945849971,1.2153651182312464,17.6938434962443,1.1058652185152686,37.3725188861857,1.1678912151933032
-model,execution model,problem type,speedup_1@1,efficiency_1@1,speedup_2@1,efficiency_2@1,speedup_4@1,efficiency_4@1,speedup_8@1,efficiency_8@1,speedup_16@1,efficiency_16@1,speedup_32@1,efficiency_32@1
 GPT-3.5,kokkos,dense_la,0.3748733387365133,0.3748733387365133,0.6835610873325082,0.3417805436662541,1.1996674349832994,0.29991685874582485,2.334455228690344,0.291806903586293,4.707111594254605,0.2941944746409128,8.383462228402035,0.2619831946375636
 GPT-3.5,kokkos,geometry,0.3012979727822783,0.3012979727822783,0.3642080967413508,0.1821040483706754,0.5760957559578642,0.14402393898946605,1.0366038331417795,0.12957547914272244,1.9605502544770208,0.1225343909048138,3.668538518001327,0.11464182868754147
 GPT-3.5,kokkos,graph,0.7412395326650871,0.7412395326650871,0.9680207190028367,0.48401035950141835,1.5157893172634915,0.3789473293158729,2.44191731303119,0.30523966412889875,3.9875336703380206,0.2492208543961263,6.649929389329907,0.2078102934165596
@@ -33,7 +30,6 @@ GPT-3.5,kokkos,sort,4.985551076169571,4.985551076169571,5.010649432352911,2.5053
 GPT-3.5,kokkos,sparse_la,0.0004449960962215412,0.0004449960962215412,0.0008734239749460548,0.0004367119874730274,0.0017341272366519888,0.0004335318091629972,0.0033644185128254337,0.0004205523141031792,0.006748107118098776,0.0004217566948811735,0.012917277242730214,0.0004036649138353192
 GPT-3.5,kokkos,stencil,0.4752402242703246,0.4752402242703246,0.9500356291712795,0.47501781458563974,1.8945822677395054,0.47364556693487636,3.7791184639646636,0.47238980799558294,7.630186334033743,0.47688664587710894,15.052913943209472,0.470403560725296
 GPT-3.5,kokkos,transform,1.2136049341447268,1.2136049341447268,2.4163336012647028,1.2081668006323514,4.823800201172887,1.2059500502932217,9.609276694512442,1.2011595868140552,19.476857209072666,1.2173035755670416,38.899404197448064,1.215606381170252
-model,execution model,problem type,speedup_1@1,efficiency_1@1,speedup_2@1,efficiency_2@1,speedup_4@1,efficiency_4@1,speedup_8@1,efficiency_8@1,speedup_16@1,efficiency_16@1,speedup_32@1,efficiency_32@1
 GPT-4,kokkos,dense_la,0.3569444899083216,0.3569444899083216,0.7316626517058835,0.36583132585294176,1.4410217756582706,0.36025544391456765,2.8516525430320723,0.35645656787900903,5.725294585495723,0.3578309115934827,10.395536645386432,0.324860520168326
 GPT-4,kokkos,fft,0.18064348366036254,0.18064348366036254,0.22023644648492374,0.11011822324246187,0.29737651145346233,0.07434412786336558,0.4222872663008739,0.052785908287609234,0.6717701778698616,0.04198563611686635,1.1136304903933616,0.03480095282479255
 GPT-4,kokkos,geometry,0.3567406690507389,0.3567406690507389,0.45923198675039334,0.22961599337519667,0.7691798660004991,0.1922949665001248,1.420008608937712,0.177501076117214,2.7719989394768105,0.17324993371730066,5.457381190958905,0.1705431622174658
@@ -46,7 +42,6 @@ GPT-4,kokkos,sort,2.647653768267662,2.647653768267662,2.679572960334517,1.339786
 GPT-4,kokkos,sparse_la,0.4749675463732183,0.4749675463732183,0.72546020665437,0.362730103327185,1.2382589450887436,0.3095647362721859,2.2268310675050595,0.27835388343813244,4.0125540198143534,0.2507846262383971,6.863586733520972,0.21448708542253037
 GPT-4,kokkos,stencil,0.5069368419791573,0.5069368419791573,1.023166164790775,0.5115830823953875,2.055311680864417,0.5138279202161042,4.078505173990893,0.5098131467488616,8.088955008298928,0.505559688018683,15.7437846086328,0.491993269019775
 GPT-4,kokkos,transform,1.2801713487352007,1.2801713487352007,2.5678133530232716,1.2839066765116358,5.185280713937532,1.296320178484383,10.2915685863688,1.2864460732961,20.586818880872205,1.2866761800545128,40.39788183100453,1.2624338072188916
-model,execution model,problem type,speedup_1@1,efficiency_1@1,speedup_2@1,efficiency_2@1,speedup_4@1,efficiency_4@1,speedup_8@1,efficiency_8@1,speedup_16@1,efficiency_16@1,speedup_32@1,efficiency_32@1
 Phind-V2,kokkos,dense_la,0.4710271957025354,0.4710271957025354,0.8332218394634384,0.4166109197317192,1.5529477903871365,0.3882369475967841,2.981745484280533,0.37271818553506664,5.823216690082931,0.36395104313018317,10.61504332117168,0.331720103786615
 Phind-V2,kokkos,geometry,0.3942860528945495,0.3942860528945495,0.5067378232396055,0.25336891161980274,0.8454717230993197,0.21136793077482993,1.5587432720890384,0.1948429090111298,3.007036631792662,0.18793978948704138,5.987617987558664,0.18711306211120826
 Phind-V2,kokkos,graph,0.5983573975848678,0.5983573975848678,1.0861886941957823,0.5430943470978912,1.6905981768861762,0.42264954422154405,3.0319360214770334,0.3789920026846292,5.658203967795536,0.353637747987221,10.71088040281966,0.33471501258811437
@@ -58,7 +53,6 @@ Phind-V2,kokkos,sort,0.09521139911253719,0.09521139911253719,0.19053076805566296
 Phind-V2,kokkos,sparse_la,0.000563788375229045,0.000563788375229045,0.0011268322404455822,0.0005634161202227911,0.0022517817491172105,0.0005629454372793026,0.00449235765144036,0.000561544706430045,0.00897960034614743,0.0005612250216342143,0.017951439813263387,0.0005609824941644808
 Phind-V2,kokkos,stencil,0.5458428013892959,0.5458428013892959,1.0900270410706556,0.5450135205353278,2.1610951027877294,0.5402737756969324,4.27692384340353,0.5346154804254413,8.710684927394228,0.5444178079621392,17.25511908594886,0.5392224714359019
 Phind-V2,kokkos,transform,1.251893340981431,1.251893340981431,2.501497357104623,1.2507486785523114,4.988430518358764,1.247107629589691,9.918968638437567,1.239871079804696,19.399915415186335,1.212494713449146,36.63408153531702,1.144815047978657
-model,execution model,problem type,speedup_1@1,efficiency_1@1,speedup_2@1,efficiency_2@1,speedup_4@1,efficiency_4@1,speedup_8@1,efficiency_8@1,speedup_16@1,efficiency_16@1,speedup_32@1,efficiency_32@1
 StarCoderBase,kokkos,dense_la,0.47012694614931916,0.47012694614931916,0.930899710057654,0.465449855028827,1.8335436349190986,0.45838590872977464,3.6385382600941063,0.4548172825117633,7.181805765585114,0.44886286034906964,13.284162619404947,0.4151300818564046
 StarCoderBase,kokkos,geometry,0.46467064414294024,0.46467064414294024,0.47397187100297966,0.23698593550148983,0.4727359728670373,0.11818399321675932,0.4770613445563825,0.05963266806954781,0.477752569969225,0.029859535623076562,0.4780203783517691,0.014938136823492784
 StarCoderBase,kokkos,graph,0.6497844624716448,0.6497844624716448,0.8202551660892956,0.4101275830446478,1.1066777312875538,0.27666943282188844,1.6093196468360351,0.2011649558545044,2.548313180327052,0.15926957377044074,4.0972319155504,0.12803849736095