LTR读书笔记.docxVIP

LTR读书笔记配置1.1 路径ltr/ml/conf1.2 内容[DEFAULT]IN_DATA:输入数据DATA_PATH:数据路径TRAIN_EXT:训练集TEST_EXT:测试集[global_config]tool_num:调用工具数量run_flow:产品线flow_zhidao:工具顺序；如flow_zhidao = split_data, trans_data_train,trans_data_test, train_svmrank, test_svmrank, testeval_test[id:13]tool_name = split_datatool_addr = tools/RankDataExtractionProcess.shsplit_para = 1,1in_data = %(IN_DATA)strain_data = %(DATA_PATH)s%(TRAIN_EXT)stest_data = %(DATA_PATH)s%(TEST_EXT)s1.3 工具1.3.1 sample_dataholmes/data_sample/batch.sh1.3.2 train_flowholmes/rerank/max-entropy-holmes.1.0/bin.new/main.py1.3.3 data2svm_train_classholmes/data2svm/main.py1.3.4 data2svm_test_classholmes/data2svm/main.py1.3.5 train_maxeholmes/train_maxe/main.py1.3.6 test_maxeholmes/test_maxe/main.py1.3.7 train_svmrank训练集common_lib/train.pyholmes/conf/rank_svm.cfg#!./third_party_lib/python/bin/pythonimport sysimportConfigParserimportsubprocessimportosimportgetoptimporttempfilefromcommon_lib.config import *deftrain_basicwei(train_file, config, model):return fake basci_wei traindeftrain_svm(train_file, config, model_file):options = config.items(svm)options.append((training_file,train_file))options.append((model_out, model_file))cmds = [] cmds.append(os.getcwd() + os.sep + SVM_TRAIN_PATH)for name, value in options:cmds.append(-- + name)cmds.append(value)printsubprocess.check_output(cmds) deftrain_svm_ori(train_file, config, model_file):options = config.items(svm)options.append((training_file,train_file))options.append((model_out, model_file))cmds = [] cmds.append(os.getcwd() + os.sep + SVM_ORI_TRAIN_PATH)for name, value in options:cmds.append(-- + name)cmds.append(value)printsubprocess.check_output(cmds) RF_MODEL_FILE = rf_model.confdeftrain_rf(train_file, config, model_file):options = config.items(rf)ifos.path.isabs(train_file):dir = os.path.dirname(train_file) else:dir = os.path.dirname(os.getcwd() + os.sep + train_file)model_conf = dir + os.sep + RF_MODEL_FILE MC = open(model_conf, w)print MC, train_data :, os.path.basename(train