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: 04f98ffa5e
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '04f98ffa5e'
${ noResults }
learnware/examples/dataset_table_workflow/config.py

298 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, 500, 1000, 2000]

  7. hetero_n_repeat_list = [10, 10, 10, 10, 10, 10, 3, 3, 3]

  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.     "M5": {

  65.         "lgb": {

  66.             "params": {

  67.                 "boosting_type": "gbdt",

  68.                 "objective": "rmse",

  69.                 "metric": "rmse",

  70.                 "learning_rate": 0.015,

  71.                 "num_leaves": 300,

  72.                 "max_depth": 500,

  73.                 "n_estimators": 100000,

  74.                 "boost_from_average": False,

  75.                 "num_threads": 32,

  76.                 "verbose": -1

  77.             },

  78.             "MAX_ROUNDS": 1000,

  79.             "early_stopping_rounds": 1000

  80.         }

  81.     },

  82.     "PFS": {

  83.         "lgb": {

  84.             "params": {

  85.                 "boosting_type": "gbdt",

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

  87.                 "learning_rate": 0.01,

  88.                 "objective": "rmse",

  89.                 "metric": "rmse",

  90.                 "feature_fraction": 0.75,

  91.                 "bagging_fraction": 0.75,

  92.                 "bagging_freq": 5,

  93.                 "seed": 1,

  94.                 "verbose": -100,

  95.                 "n_estimators": 100000,

  96.             },

  97.             "MAX_ROUNDS": 1000,

  98.             "early_stopping_rounds": 1000

  99.         }

  100.     },

  101.     "PFS_HOMO": {

  102.         "lgb": {

  103.             "params": {

  104.                 "boosting_type": "gbdt",

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

  106.                 "learning_rate": 0.01,

  107.                 "objective": "rmse",

  108.                 "metric": "rmse",

  109.                 "feature_fraction": 0.75,

  110.                 "bagging_fraction": 0.75,

  111.                 "bagging_freq": 5,

  112.                 "seed": 1,

  113.                 "verbose": -100,

  114.                 "n_estimators": 100000,

  115.             },

  116.             "MAX_ROUNDS": 1000,

  117.             "early_stopping_rounds": 1000

  118.         }

  119.     }

  120. }

  121. 

  122. homo_table_benchmark_config = BenchmarkConfig(

  123.     name="PFS_HOMO",

  124.     user_num=53,

  125.     learnware_ids=[

  126.         "00002265", "00002266", "00002267", "00002268",

  127.         "00002269", "00002270", "00002271", "00002272",

  128.         "00002273", "00002274", "00002275", "00002276",

  129.         "00002277", "00002278", "00002279", "00002280",

  130.         "00002281", "00002282", "00002283", "00002284",

  131.         "00002285", "00002286", "00002287", "00002288",

  132.         "00002289", "00002290", "00002291", "00002292",

  133.         "00002293", "00002294", "00002295", "00002296",

  134.         "00002297", "00002298", "00002299", "00002300",

  135.         "00002301", "00002302", "00002303", "00002304",

  136.         "00002305", "00002306", "00002307", "00002308",

  137.         "00002309", "00002310", "00002311", "00002312",

  138.         "00002313", "00002314", "00002315", "00002316",

  139.         "00002317"

  140.     ],

  141.     test_data_path="PFS_HOMO/test_data.zip",

  142.     train_data_path="PFS_HOMO/train_data.zip",

  143.     extra_info_path="PFS_HOMO/extra_info.zip"

  144. )

  145. 

  146. hetero_cross_feat_eng_benchmark_config = BenchmarkConfig(  

  147.     name="PFS",

  148.     user_num=41,

  149.     learnware_ids = [

  150.         "00000342", "00000343", "00000344", "00000345",

  151.         "00000346", "00000347", "00000348", "00000349",

  152.         "00000350", "00000351", "00000352", "00000353",

  153.         "00000354", "00000355", "00000356", "00000357",

  154.         "00000358", "00000359", "00000360", "00000361",

  155.         "00000362", "00000363", "00000364", "00000365",

  156.         "00000366", "00000367", "00000368", "00000369",

  157.         "00000370", "00000371", "00000372", "00000373",

  158.         "00000374", "00000375", "00000376", "00000377",

  159.         "00000378", "00000379", "00000380", "00000381",

  160.         "00000382", "00000383", "00000384", "00000385",

  161.         "00000386", "00000387", "00000388", "00000389",

  162.         "00000390", "00000391", "00000392", "00000393",

  163.         "00000394", "00000395", "00000396", "00000397",

  164.         "00000398", "00000399", "00000400", "00000401",

  165.         "00000402", "00000403", "00000404", "00000405",

  166.         "00000406", "00000407", "00000408", "00000409",

  167.         "00000410", "00000411", "00000412", "00000413",

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

  169.         "00000418", "00000419", "00000420", "00000421",

  170.         "00000422", "00000423", "00000424", "00000425",

  171.         "00000426", "00000427", "00000428", "00000429",

  172.         "00000430", "00000431", "00000432", "00000433",

  173.         "00000434", "00000435", "00000436", "00000437",

  174.         "00000438", "00000439", "00000440", "00000441",

  175.         "00000442", "00000443", "00000444", "00000730",

  176.         "00000731", "00000732", "00000733", "00000734",

  177.         "00000735", "00000736", "00000737", "00000738",

  178.         "00000739", "00000740", "00000741", "00000742",

  179.         "00000743", "00000744", "00000745", "00000746",

  180.         "00000747", "00000748", "00000749", "00000750",

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

  182.         "00000755", "00000756", "00000757", "00000758",

  183.         "00000759", "00000760", "00000761", "00000762",

  184.         "00000763", "00000764", "00000765", "00000766",

  185.         "00000767", "00000768", "00000769", "00000770",

  186.         "00000771", "00000772", "00000773", "00000774",

  187.         "00000775", "00000776", "00000777", "00000778",

  188.         "00000779", "00000780", "00000781", "00000782",

  189.         "00000783", "00000786", "00000787", "00000788",

  190.         "00000789", "00000790", "00000791", "00000792",

  191.         "00000793", "00000794", "00000795", "00000796",

  192.         "00000797", "00000798", "00000799", "00000800",

  193.         "00000801", "00000802", "00000803", "00000804",

  194.         "00000805", "00000806", "00000807", "00000808",

  195.         "00000809", "00000810", "00000811", "00000812",

  196.         "00000813", "00000814", "00000815", "00000816",

  197.         "00000817", "00000818", "00000819", "00000820",

  198.         "00000821", "00000822", "00000823", "00000824",

  199.         "00000825", "00000826", "00000827", "00000828",

  200.         "00000829", "00000830", "00000831", "00000832",

  201.         "00000833", "00000834", "00000835", "00000836",

  202.         "00000837", "00000838", "00000839", "00000859",

  203.         "00000860", "00000861", "00000862", "00000863",

  204.         "00000864", "00000865", "00000866", "00000867",

  205.         "00000868", "00000869", "00000870", "00000871",

  206.         "00000872", "00000873", "00000874", "00000875",

  207.         "00000876", "00000877", "00000878", "00000879",

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

  209.         "00000884", "00000885", "00000886", "00000887",

  210.         "00000888", "00000889", "00000890", "00000891",

  211.         "00000892", "00000893", "00000894", "00000895",

  212.         "00000896", "00000897", "00000898", "00000899",

  213.         "00000900", "00000901", "00000902", "00000903",

  214.         "00000904", "00000905", "00000906", "00000907",

  215.         "00000908", "00000909", "00000910", "00000911",

  216.         "00000912"

  217.     ],

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

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

  220.     extra_info_path="PFS/extra_info.zip"

  221. )

  222. 

  223. hetero_cross_task_benchmark_config = BenchmarkConfig(

  224.     name="M5",

  225.     user_num=30,

  226.     learnware_ids = [

  227.         "00000342", "00000343", "00000344", "00000345",

  228.         "00000346", "00000347", "00000348", "00000349",

  229.         "00000350", "00000351", "00000352", "00000353",

  230.         "00000354", "00000355", "00000356", "00000357",

  231.         "00000358", "00000359", "00000360", "00000361",

  232.         "00000362", "00000363", "00000364", "00000365",

  233.         "00000366", "00000367", "00000368", "00000369",

  234.         "00000370", "00000371", "00000372", "00000373",

  235.         "00000374", "00000375", "00000376", "00000377",

  236.         "00000378", "00000379", "00000380", "00000381",

  237.         "00000382", "00000383", "00000384", "00000385",

  238.         "00000386", "00000387", "00000388", "00000389",

  239.         "00000390", "00000391", "00000392", "00000393",

  240.         "00000394", "00000395", "00000396", "00000397",

  241.         "00000398", "00000399", "00000400", "00000401",

  242.         "00000402", "00000403", "00000404", "00000405",

  243.         "00000406", "00000407", "00000408", "00000409",

  244.         "00000410", "00000411", "00000412", "00000413",

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

  246.         "00000418", "00000419", "00000420", "00000421",

  247.         "00000422", "00000423", "00000424", "00000425",

  248.         "00000426", "00000427", "00000428", "00000429",

  249.         "00000430", "00000431", "00000432", "00000433",

  250.         "00000434", "00000435", "00000436", "00000437",

  251.         "00000438", "00000439", "00000440", "00000441",

  252.         "00000442", "00000443", "00000444", "00000730",

  253.         "00000731", "00000732", "00000733", "00000734",

  254.         "00000735", "00000736", "00000737", "00000738",

  255.         "00000739", "00000740", "00000741", "00000742",

  256.         "00000743", "00000744", "00000745", "00000746",

  257.         "00000747", "00000748", "00000749", "00000750",

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

  259.         "00000755", "00000756", "00000757", "00000758",

  260.         "00000759", "00000760", "00000761", "00000762",

  261.         "00000763", "00000764", "00000765", "00000766",

  262.         "00000767", "00000768", "00000769", "00000770",

  263.         "00000771", "00000772", "00000773", "00000774",

  264.         "00000775", "00000776", "00000777", "00000778",

  265.         "00000779", "00000780", "00000781", "00000782",

  266.         "00000783", "00000786", "00000787", "00000788",

  267.         "00000789", "00000790", "00000791", "00000792",

  268.         "00000793", "00000794", "00000795", "00000796",

  269.         "00000797", "00000798", "00000799", "00000800",

  270.         "00000801", "00000802", "00000803", "00000804",

  271.         "00000805", "00000806", "00000807", "00000808",

  272.         "00000809", "00000810", "00000811", "00000812",

  273.         "00000813", "00000814", "00000815", "00000816",

  274.         "00000817", "00000818", "00000819", "00000820",

  275.         "00000821", "00000822", "00000823", "00000824",

  276.         "00000825", "00000826", "00000827", "00000828",

  277.         "00000829", "00000830", "00000831", "00000832",

  278.         "00000833", "00000834", "00000835", "00000836",

  279.         "00000837", "00000838", "00000839", "00000859",

  280.         "00000860", "00000861", "00000862", "00000863",

  281.         "00000864", "00000865", "00000866", "00000867",

  282.         "00000868", "00000869", "00000870", "00000871",

  283.         "00000872", "00000873", "00000874", "00000875",

  284.         "00000876", "00000877", "00000878", "00000879",

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

  286.         "00000884", "00000885", "00000886", "00000887",

  287.         "00000888", "00000889", "00000890", "00000891",

  288.         "00000892", "00000893", "00000894", "00000895",

  289.         "00000896", "00000897", "00000898", "00000899",

  290.         "00000900", "00000901", "00000902", "00000903",

  291.         "00000904", "00000905", "00000906", "00000907",

  292.         "00000908", "00000909", "00000910", "00000911",

  293.         "00000912"

  294.     ],

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

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

  297.     extra_info_path="M5/extra_info.zip"

  298. )