2019-08-21工作进展

1. 使用测试集看一下效果
   hs_test_data_dssm_v2_5 : | se_keyword_mainse_ws | title_mainse_ws | label |
   hs_tmp_267 : | query_id | query_ws | video_id | video_ws |

drop table hs_tmp_272;
yes
create table hs_tmp_272 as select row_number()over() as query_id, se_keyword_mainse_ws as query_ws, 20 as video_id, title_mainse_ws as video_ws, label from hs_test_data_dssm_v2_6;

pai -name tensorflow140 -Dscript="file:///home/hengsong/origin_deep_cluster_odps_8.tar.gz" -DentryFile="inference_v8.py" -Dcluster='{"worker":{"count":50, "cpu":200, "memory":4000}, "ps":{"count":10, "cpu":200, "memory":5000}}' -DuseSparseClusterSchema=True -DenableDynamicCluster=True -Dtables="odps://graph_embedding/tables/hs_train_data_dssm_v2_6,odps://graph_embedding/tables/hs_test_data_dssm_v2_6,odps://graph_embedding/tables/hs_tmp_272" -Doutputs="odps://graph_embedding/tables/hs_dssm_result_3" -DcheckpointDir="oss://bucket-automl/hengsong/?role_arn=acs:ram::1293303983251548:role/graph2018&host=cn-hangzhou.oss-internal.aliyun-inc.com" -DuserDefinedParameters="--learning_rate=3e-4 --batch_size=1024 --is_save_model=True --attention_type=1 --num_epochs=1 --ckpt=hs_ugc_video_4e_8.ckpt-3" -DuseSparseClusterSchema=True;

create table graph_embedding.hs_tmp_273 as
select cast(a.label as bigint) as label, cast(b.score as bigint) as score from
(select * from graph_embedding.hs_tmp_272)a join (select * from graph_embedding.hs_dssm_result_3)b on a.query_id == b.query_id;

测试集效果：
训练集效果

tunnel download -fd '|' graph_embedding.hs_tmp_273 /home/hengsong/hs_label_score.txt;