serving one model multiple versions error

[mattxia]

servableHandle = {ServableHandle@9593} "ServableHandle(UntypedServableHandle({name: lr, version: 6},com.tencent.angel.serving.core.SimpleLoader@7849b420))" untypedHandle = {UntypedServableHandle@9611} "UntypedServableHandle({name: lr, version: 6},com.tencent.angel.serving.core.SimpleLoader@7849b420)" id = {ServableId@9613} "{name: lr, version: 6}" name = "lr" version = 6 loader = {SimpleLoader@9614} creator = {AngelSourceAdapter$$anonfun$1@9618} "" postLoadResourceEstimate = {AngelSourceAdapter$$anonfun$3@9619} "" memorizedResourceEstimator = {ResourceAllocation@9620} "resource_quantities {\n resource {\n device: "CPU"\n device_instance {\n }\n kind: "Memmory"\n }\n}\n" resourceUtil = {ResourceUtil@9621} servable_ = {AngelSavedModelBundle@6931} model = {LocalModel@6963} conf = {SharedConf@7028} "{"ml.reg.l1":"0.0","angel.load.model.path":"D:\prj\angel\serving-develop\serving-develop\models\angel\lr\lr-model\3","ml.batch.sample.ratio":"1.0","ml.optimizer.json.provider":"com.tencent.angel.ml.core.local.LocalOptimizerProvider","ml.epoch.num":"10","ml.feature.index.range":"123","ml.data.validate.ratio":"0.1","ml.data.type":"libsvm","ml.learn.rate":"0.5","ml.json.conf.file":### "D:\prj\angel\serving-develop\serving-develop\models\angel\lr\lr-model\3\graph.json","ml.opt.adam.gamma":"0.99","ml.model.type":"T_FLOAT_DENSE","ml.opt.adam.beta":"0.9","ml.reg.l2":"0.0","ml.model.size":"-1","ml.simpleinputlayer.matrix.output.format":"com.tencent.angel.model.output.format.ColIdValueTextRowFormat","graphJson":{"input":{"name":"input","type":"simpleinputlayer","outputdim":1,"transfunc":"identity","optimizer":{"type":"adam","reg2":0.01}},"simplelosslayer":{"name":"simplelosslayer","type":"simplelosslayer","lossfunc":"logloss","inputlayer":"input"}}}" placeHolder = {PlaceHolder@7029} variableManager = {LocalVariableManager@7030} variableProvider = {LocalVariableProvider@7031} graph = {Graph@7024} Method threw 'scala.MatchError' exception. Cannot evaluate com.tencent.angel.ml.core.network.Graph.toString() dataFormat = "libsvm" indexRange = 123 validIndexNum = 123 modelType = {RowType@7033} "T_FLOAT_DENSE" isSparseFormat = true LOG = {Log4jLoggerAdapter@6940} "org.slf4j.impl.Log4jLoggerAdapter(com.tencent.angel.serving.servables.angel.AngelSavedModelBundle)"

---

9/04/18 20:23:14 INFO common.ServiceImpl$ : servableHandle {name: lr, version: 6} 19/04/18 20:23:17 INFO angel.AngelSavedModelBundle : runPredict:
19/04/18 20:23:17 INFO metrics.MetricsCollector : metric_name="PredictMetric", metric_version=1, model_name="lr", model_version=6, is_success=false, predict_time_ms=6766 19/04/18 20:23:37 INFO metrics.MetricsCollector : metric_name="PredictSummary", model_name="lr", model_version=6, prediction_count_total=1, prediction_count_success=0, prediction_count_failed=1, total_predict_time_ms=0, count_distribution0=0, count_distribution1=0, count_distribution2=0, count_distribution3=0 19/04/18 20:24:08 INFO metrics.MetricsCollector : metric_name="PredictSummary", model_name="lr", model_version=6, prediction_count_total=1, prediction_count_success=0, prediction_count_failed=1, total_predict_time_ms=0, count_distribution0=0, count_distribution1=0, count_distribution2=0, count_distribution3=0 19/04/18 20:24:08 INFO common.ServiceImpl$ : modelSpec: lr, value: 3 19/04/18 20:24:08 INFO common.ServiceImpl$ : servableHandle {name: lr, version: 3} 19/04/18 20:24:09 INFO angel.AngelSavedModelBundle : runPredict:
四月 18, 2019 8:24:09 下午 com.github.fommil.netlib.BLAS 警告: Failed to load implementation from: com.github.fommil.netlib.NativeSystemBLAS 四月 18, 2019 8:24:09 下午 com.github.fommil.jni.JniLoader liberalLoad 信息: successfully loaded C:\Users\xxx~1\AppData\Local\Temp\jniloader1438167409672823982netlib-native_ref-win-x86_64.dll

1.there are one model named lr, but two version, 3 and 6, in the model folder 2.policy all or specific or latest with num versions = 2 3.the latest loaded version(normally 3)will work, but the earlier one won't work. 4.event the netlib isn't triggered to be loaded. 5.the 6 version conf shows the 3 version info.

[wangcaihua]

I got it.

Since the "SharedConf" is single, that cause problem. We will remove some limitation to enable multiple model versions of the Angel.