beimingwu
/
learnware
Not watched
1
0
Fork 0
Code
Releases 1 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
Tree: ed0c4bd7fd
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'ed0c4bd7fd'
${ noResults }
learnware/examples/dataset_table_workflow/config.py

296 lines
13 kB
Raw Blame History 

  1. from learnware.tests.benchmarks import BenchmarkConfig

  2. 

  3. 

  4. homo_n_labeled_list = [100, 200, 500, 1000, 2000, 4000, 6000, 8000, 10000]

  5. homo_n_repeat_list = [10, 10, 10, 3, 3, 3, 3, 3, 3]

  6. hetero_n_labeled_list = [10, 30, 50, 75, 100, 200]

  7. hetero_n_repeat_list = [10, 10, 10, 10, 10, 10]

  8. 

  9. 

  10. user_semantic = {

  11.     "Data": {"Values": ["Table"], "Type": "Class"},

  12.     "Task": {"Values": ["Regression"], "Type": "Class"},

  13.     "Library": {"Values": ["Others"], "Type": "Class"},

  14.     "Scenario": {"Values": ["Business"], "Type": "Tag"},

  15.     "Description": {"Values": "", "Type": "String"},

  16.     "Name": {"Values": "", "Type": "String"},

  17. }

  18. 

  19. styles = {

  20.     'user_model': {"color": "navy", "marker": "o", "linestyle": "-"},  

  21.     'select_score': {'color': 'gold', 'marker': 's', 'linestyle': '--'},  

  22.     'oracle_score': {'color': 'darkorange', 'marker': '^', 'linestyle': '-.'},  

  23.     'mean_score': {'color': 'gray', 'marker': 'x', 'linestyle': ':'},

  24.     'single_aug': {'color': 'gold', 'marker': 's', 'linestyle': '--'},

  25.     'multiple_avg': {'color': 'blue', 'marker': '*', 'linestyle': '-'},  

  26.     'multiple_aug': {'color': 'purple', 'marker': 'd', 'linestyle': '--'},  

  27.     'ensemble_pruning': {"color": "magenta", "marker": "d", "linestyle": "-."} 

  28. }

  29. 

  30. labels = {

  31.     'user_model': "User Model",  

  32.     'single_aug': "Single Learnware Reuse (FeatAug)",

  33.     "select_score": "Single Learnware Reuse (FeatAug)",

  34.     'multiple_aug': "Multiple Learnware Reuse (FeatAug)",

  35.     'ensemble_pruning': "Multiple Learnware Reuse (EnsemblePrune)",

  36.     'multiple_avg': "Multiple Learnware Reuse (Averaging)"

  37. }

  38. 

  39. align_model_params = {

  40.     "network_type": "ArbitraryMapping",  # ["ArbitraryMapping", "BaseMapping", "BaseMapping_BN", "BaseMapping_Dropout"]

  41.     "num_epoch": 50,

  42.     "lr": 1e-5,

  43.     "dropout_ratio": 0.2,

  44.     "activation": "relu",

  45.     "use_bn": True,

  46.     "hidden_dims": [128, 256, 128, 256],

  47. }

  48. 

  49. market_mapping_params = {

  50.     "lr": 1e-4, 

  51.     "num_epoch": 50,

  52.     "batch_size": 64,  

  53.     "num_partition": 2,  # num of column partitions for pos/neg sampling

  54.     "overlap_ratio": 0.7,  # specify the overlap ratio of column partitions during the CL

  55.     "hidden_dim": 256,  # the dimension of hidden embeddings

  56.     "num_layer": 6,  # the number of transformer layers used in the encoder

  57.     "num_attention_head": 8,  # the numebr of heads of multihead self-attention layer in the transformers, should be divisible by hidden_dim

  58.     "hidden_dropout_prob": 0.5,  # the dropout ratio in the transformer encoder

  59.     "ffn_dim": 512,  # the dimension of feed-forward layer in the transformer layer

  60.     "activation": "leakyrelu",

  61. }

  62. 

  63. user_model_params = {

  64.     "Corporacion": {

  65.         "lgb": {

  66.             "params": {

  67.                 "num_leaves": 31,

  68.                 "objective": "regression",

  69.                 "learning_rate": 0.1,

  70.                 "feature_fraction": 0.8,

  71.                 "bagging_fraction": 0.8,

  72.                 "bagging_freq": 2,

  73.                 "metric": "l2",

  74.                 "num_threads": 4,

  75.                 "verbose": -1,

  76.             },

  77.             "MAX_ROUNDS": 500,

  78.             "early_stopping_rounds": 50,

  79.         }

  80.     },

  81.     "M5": {

  82.         "lgb": {

  83.             "params": {

  84.                 "boosting_type": "gbdt",

  85.                 "objective": "rmse",

  86.                 "metric": "rmse",

  87.                 "learning_rate": 0.015,

  88.                 "num_leaves": 300,

  89.                 "max_depth": 500,

  90.                 "n_estimators": 100000,

  91.                 "boost_from_average": False,

  92.                 "num_threads": 32,

  93.                 "verbose": -1

  94.             },

  95.             "MAX_ROUNDS": 1000,

  96.             "early_stopping_rounds": 1000

  97.         }

  98.     },

  99.     "PFS": {

  100.         "lgb": {

  101.             "params": {

  102.                 "boosting_type": "gbdt",

  103.                 "num_leaves": 2**7 - 1,

  104.                 "learning_rate": 0.01,

  105.                 "objective": "rmse",

  106.                 "metric": "rmse",

  107.                 "feature_fraction": 0.75,

  108.                 "bagging_fraction": 0.75,

  109.                 "bagging_freq": 5,

  110.                 "seed": 1,

  111.                 "verbose": -100,

  112.                 "n_estimators": 100000,

  113.             },

  114.             "MAX_ROUNDS": 1000,

  115.             "early_stopping_rounds": 1000

  116.         }

  117.     }

  118. }

  119. 

  120. homo_table_benchmark_config = BenchmarkConfig(

  121.     name="Corporacion",

  122.     user_num=54,

  123.     learnware_ids=[

  124.         "00000912", "00000911", "00000910", "00000909",

  125.         "00000908", "00000907", "00000906", "00000905",

  126.         "00000904", "00000903", "00000902", "00000901",

  127.         "00000900", "00000899", "00000898", "00000897",

  128.         "00000896", "00000895", "00000894", "00000893",

  129.         "00000892", "00000891", "00000890", "00000889",

  130.         "00000888", "00000887", "00000886", "00000885",

  131.         "00000884", "00000883", "00000882", "00000881",

  132.         "00000880", "00000879", "00000878", "00000877",

  133.         "00000876", "00000875", "00000874", "00000873",

  134.         "00000872", "00000871", "00000870", "00000869",

  135.         "00000868", "00000867", "00000866", "00000865",

  136.         "00000864", "00000863", "00000862", "00000861",

  137.         "00000860", "00000859"

  138.     ],

  139.     test_data_path="Corporacion/test_data.zip",

  140.     train_data_path="Corporacion/train_data.zip",

  141.     extra_info_path="Corporacion/extra_info.zip"

  142. )

  143. 

  144. hetero_cross_feat_eng_benchmark_config = BenchmarkConfig(  

  145.     name="PFS",

  146.     user_num=41,

  147.     learnware_ids = [

  148.         "00000394", "00000393", "00000392", "00000391",

  149.         "00000390", "00000389", "00000388", "00000387",

  150.         "00000386", "00000385", "00000384", "00000383",

  151.         "00000382", "00000381", "00000380", "00000379",

  152.         "00000378", "00000377", "00000376", "00000375",

  153.         "00000374", "00000373", "00000372", "00000371",

  154.         "00000370", "00000369", "00000368", "00000367",

  155.         "00000366", "00000365", "00000364", "00000363",

  156.         "00000362", "00000361", "00000360", "00000359",

  157.         "00000358", "00000357", "00000356", "00000355",

  158.         "00000354", "00000353", "00000352", "00000351",

  159.         "00000350", "00000349", "00000348", "00000347",

  160.         "00000346", "00000345", "00000344", "00000343",

  161.         "00000342", "00000444", "00000443", "00000442",

  162.         "00000441", "00000440", "00000439", "00000438",

  163.         "00000437", "00000436", "00000435", "00000434",

  164.         "00000433", "00000432", "00000431", "00000430",

  165.         "00000429", "00000428", "00000427", "00000426",

  166.         "00000425", "00000424", "00000423", "00000422",

  167.         "00000421", "00000420", "00000419", "00000418",

  168.         "00000417", "00000416", "00000415", "00000414",

  169.         "00000413", "00000412", "00000411", "00000410",

  170.         "00000409", "00000408", "00000407", "00000406",

  171.         "00000405", "00000404", "00000403", "00000402",

  172.         "00000401", "00000400", "00000399", "00000398",

  173.         "00000397", "00000396", "00000395", "00000783",

  174.         "00000782", "00000781", "00000780", "00000779",

  175.         "00000778", "00000777", "00000776", "00000775",

  176.         "00000774", "00000773", "00000772", "00000771",

  177.         "00000770", "00000769", "00000768", "00000767",

  178.         "00000766", "00000765", "00000764", "00000763",

  179.         "00000762", "00000761", "00000760", "00000759",

  180.         "00000758", "00000757", "00000756", "00000755",

  181.         "00000754", "00000753", "00000752", "00000751",

  182.         "00000750", "00000749", "00000748", "00000747",

  183.         "00000746", "00000745", "00000744", "00000743",

  184.         "00000742", "00000741", "00000740", "00000739",

  185.         "00000738", "00000737", "00000736", "00000735",

  186.         "00000734", "00000733", "00000732", "00000731",

  187.         "00000730", "00000839", "00000838", "00000837",

  188.         "00000836", "00000835", "00000834", "00000833",

  189.         "00000832", "00000831", "00000830", "00000829",

  190.         "00000828", "00000827", "00000826", "00000825",

  191.         "00000824", "00000823", "00000822", "00000821",

  192.         "00000820", "00000819", "00000818", "00000817",

  193.         "00000816", "00000815", "00000814", "00000813",

  194.         "00000812", "00000811", "00000810", "00000809",

  195.         "00000808", "00000807", "00000806", "00000805",

  196.         "00000804", "00000803", "00000802", "00000801",

  197.         "00000800", "00000799", "00000798", "00000797",

  198.         "00000796", "00000795", "00000794", "00000793",

  199.         "00000792", "00000791", "00000790", "00000789",

  200.         "00000788", "00000787", "00000786", "00000912",

  201.         "00000911", "00000910", "00000909", "00000908",

  202.         "00000907", "00000906", "00000905", "00000904",

  203.         "00000903", "00000902", "00000901", "00000900",

  204.         "00000899", "00000898", "00000897", "00000896",

  205.         "00000895", "00000894", "00000893", "00000892",

  206.         "00000891", "00000890", "00000889", "00000888",

  207.         "00000887", "00000886", "00000885", "00000884",

  208.         "00000883", "00000882", "00000881", "00000880",

  209.         "00000879", "00000878", "00000877", "00000876",

  210.         "00000875", "00000874", "00000873", "00000872",

  211.         "00000871", "00000870", "00000869", "00000868",

  212.         "00000867", "00000866", "00000865", "00000864",

  213.         "00000863", "00000862", "00000861", "00000860",

  214.         "00000859"

  215.     ],

  216.     test_data_path="PFS/test_data.zip",

  217.     train_data_path="PFS/train_data.zip",

  218.     extra_info_path="PFS/extra_info.zip"

  219. )

  220. 

  221. hetero_cross_task_benchmark_config = BenchmarkConfig(

  222.     name="M5",

  223.     user_num=9,

  224.     learnware_ids = [

  225.         "00000394", "00000393", "00000392", "00000391",

  226.         "00000390", "00000389", "00000388", "00000387",

  227.         "00000386", "00000385", "00000384", "00000383",

  228.         "00000382", "00000381", "00000380", "00000379",

  229.         "00000378", "00000377", "00000376", "00000375",

  230.         "00000374", "00000373", "00000372", "00000371",

  231.         "00000370", "00000369", "00000368", "00000367",

  232.         "00000366", "00000365", "00000364", "00000363",

  233.         "00000362", "00000361", "00000360", "00000359",

  234.         "00000358", "00000357", "00000356", "00000355",

  235.         "00000354", "00000353", "00000352", "00000351",

  236.         "00000350", "00000349", "00000348", "00000347",

  237.         "00000346", "00000345", "00000344", "00000343",

  238.         "00000342", "00000444", "00000443", "00000442",

  239.         "00000441", "00000440", "00000439", "00000438",

  240.         "00000437", "00000436", "00000435", "00000434",

  241.         "00000433", "00000432", "00000431", "00000430",

  242.         "00000429", "00000428", "00000427", "00000426",

  243.         "00000425", "00000424", "00000423", "00000422",

  244.         "00000421", "00000420", "00000419", "00000418",

  245.         "00000417", "00000416", "00000415", "00000414",

  246.         "00000413", "00000412", "00000411", "00000410",

  247.         "00000409", "00000408", "00000407", "00000406",

  248.         "00000405", "00000404", "00000403", "00000402",

  249.         "00000401", "00000400", "00000399", "00000398",

  250.         "00000397", "00000396", "00000395", "00000783",

  251.         "00000782", "00000781", "00000780", "00000779",

  252.         "00000778", "00000777", "00000776", "00000775",

  253.         "00000774", "00000773", "00000772", "00000771",

  254.         "00000770", "00000769", "00000768", "00000767",

  255.         "00000766", "00000765", "00000764", "00000763",

  256.         "00000762", "00000761", "00000760", "00000759",

  257.         "00000758", "00000757", "00000756", "00000755",

  258.         "00000754", "00000753", "00000752", "00000751",

  259.         "00000750", "00000749", "00000748", "00000747",

  260.         "00000746", "00000745", "00000744", "00000743",

  261.         "00000742", "00000741", "00000740", "00000739",

  262.         "00000738", "00000737", "00000736", "00000735",

  263.         "00000734", "00000733", "00000732", "00000731",

  264.         "00000730", "00000839", "00000838", "00000837",

  265.         "00000836", "00000835", "00000834", "00000833",

  266.         "00000832", "00000831", "00000830", "00000829",

  267.         "00000828", "00000827", "00000826", "00000825",

  268.         "00000824", "00000823", "00000822", "00000821",

  269.         "00000820", "00000819", "00000818", "00000817",

  270.         "00000816", "00000815", "00000814", "00000813",

  271.         "00000812", "00000811", "00000810", "00000809",

  272.         "00000808", "00000807", "00000806", "00000805",

  273.         "00000804", "00000803", "00000802", "00000801",

  274.         "00000800", "00000799", "00000798", "00000797",

  275.         "00000796", "00000795", "00000794", "00000793",

  276.         "00000792", "00000791", "00000790", "00000789",

  277.         "00000788", "00000787", "00000786", "00000912",

  278.         "00000911", "00000910", "00000909", "00000908",

  279.         "00000907", "00000906", "00000905", "00000904",

  280.         "00000903", "00000902", "00000901", "00000900",

  281.         "00000899", "00000898", "00000897", "00000896",

  282.         "00000895", "00000894", "00000893", "00000892",

  283.         "00000891", "00000890", "00000889", "00000888",

  284.         "00000887", "00000886", "00000885", "00000884",

  285.         "00000883", "00000882", "00000881", "00000880",

  286.         "00000879", "00000878", "00000877", "00000876",

  287.         "00000875", "00000874", "00000873", "00000872",

  288.         "00000871", "00000870", "00000869", "00000868",

  289.         "00000867", "00000866", "00000865", "00000864",

  290.         "00000863", "00000862", "00000861", "00000860",

  291.         "00000859"

  292.     ],

  293.     test_data_path="M5/test_data.zip",

  294.     train_data_path="M5/train_data.zip",

  295.     extra_info_path="M5/extra_info.zip"

  296. )