Low Accuracy on Sentiment Analysis LSTM - GloVe20News

[Bfzanchetta]

I'm attempting to train the given Text Classifier with LSTM instead of CNN on a 8-workers BigDL's cluster. However, the training unveils a very low accuracy rate. Here's the print from the last attempt:

19/02/11 21:54:28 INFO DistriOptimizer$: [Epoch 13 15360/14923][Iteration 260][Wall Clock 597.135520654s] Epoch finished. Wall clock time is 604099.486244 ms
19/02/11 21:54:28 INFO DistriOptimizer$: [Epoch 13 15360/14923][Iteration 260][Wall Clock 597.135520654s] Validate model...
19/02/11 21:54:36 INFO DistriOptimizer$: [Epoch 13 15360/14923][Iteration 260][Wall Clock 597.135520654s] Top1Accuracy is Accuracy(correct: 226, count: 3905, accuracy: 0.05787451984635083)
19/02/11 21:54:38 INFO DistriOptimizer$: [Epoch 14 768/14923][Iteration 261][Wall Clock 605.903340897s] Trained 768 records in 1.803854653 seconds. Throughput is 425.7549 records/second. Loss is 2.9859028. 
19/02/11 21:54:40 INFO DistriOptimizer$: [Epoch 14 1536/14923][Iteration 262][Wall Clock 607.672976053s] Trained 768 records in 1.769635156 seconds. Throughput is 433.98776 records/second. Loss is 2.9682853. 
19/02/11 21:54:41 INFO DistriOptimizer$: [Epoch 14 2304/14923][Iteration 263][Wall Clock 609.393950892s] Trained 768 records in 1.720974839 seconds. Throughput is 446.2587 records/second. Loss is 2.9743903. 
19/02/11 21:54:43 INFO DistriOptimizer$: [Epoch 14 3072/14923][Iteration 264][Wall Clock 611.126353563s] Trained 768 records in 1.732402671 seconds. Throughput is 443.31494 records/second. Loss is 2.987369. 
19/02/11 21:54:45 INFO DistriOptimizer$: [Epoch 14 3840/14923][Iteration 265][Wall Clock 612.699872205s] Trained 768 records in 1.573518642 seconds. Throughput is 488.0781 records/second. Loss is 2.9755714. 
19/02/11 21:54:46 INFO DistriOptimizer$: [Epoch 14 4608/14923][Iteration 266][Wall Clock 614.428942279s] Trained 768 records in 1.729070074 seconds. Throughput is 444.1694 records/second. Loss is 2.9765773. 
19/02/11 21:54:48 INFO DistriOptimizer$: [Epoch 14 5376/14923][Iteration 267][Wall Clock 616.107890193s] Trained 768 records in 1.678947914 seconds. Throughput is 457.42932 records/second. Loss is 2.9611878. 
19/02/11 21:54:50 INFO DistriOptimizer$: [Epoch 14 6144/14923][Iteration 268][Wall Clock 617.76562979s] Trained 768 records in 1.657739597 seconds. Throughput is 463.28143 records/second. Loss is 2.9703681. 
19/02/11 21:54:51 INFO DistriOptimizer$: [Epoch 14 6912/14923][Iteration 269][Wall Clock 619.341088789s] Trained 768 records in 1.575458999 seconds. Throughput is 487.477 records/second. Loss is 2.974666. 
19/02/11 21:54:53 INFO DistriOptimizer$: [Epoch 14 7680/14923][Iteration 270][Wall Clock 621.086098227s] Trained 768 records in 1.745009438 seconds. Throughput is 440.11224 records/second. Loss is 2.976882. 
19/02/11 21:54:55 INFO DistriOptimizer$: [Epoch 14 8448/14923][Iteration 271][Wall Clock 622.767986926s] Trained 768 records in 1.681888699 seconds. Throughput is 456.62952 records/second. Loss is 2.9662917. 
19/02/11 21:54:56 INFO DistriOptimizer$: [Epoch 14 9216/14923][Iteration 272][Wall Clock 624.401519917s] Trained 768 records in 1.633532991 seconds. Throughput is 470.1466 records/second. Loss is 2.9706428. 
19/02/11 21:54:58 INFO DistriOptimizer$: [Epoch 14 9984/14923][Iteration 273][Wall Clock 626.105024646s] Trained 768 records in 1.703504729 seconds. Throughput is 450.83527 records/second. Loss is 2.9820547. 
19/02/11 21:55:00 INFO DistriOptimizer$: [Epoch 14 10752/14923][Iteration 274][Wall Clock 627.756389557s] Trained 768 records in 1.651364911 seconds. Throughput is 465.06982 records/second. Loss is 2.973231. 
19/02/11 21:55:01 INFO DistriOptimizer$: [Epoch 14 11520/14923][Iteration 275][Wall Clock 629.426095402s] Trained 768 records in 1.669705845 seconds. Throughput is 459.96124 records/second. Loss is 2.9795823. 
19/02/11 21:55:03 INFO DistriOptimizer$: [Epoch 14 12288/14923][Iteration 276][Wall Clock 631.182570675s] Trained 768 records in 1.756475273 seconds. Throughput is 437.2393 records/second. Loss is 2.9676907. 
19/02/11 21:55:05 INFO DistriOptimizer$: [Epoch 14 13056/14923][Iteration 277][Wall Clock 632.904944054s] Trained 768 records in 1.722373379 seconds. Throughput is 445.89636 records/second. Loss is 2.9707906. 
19/02/11 21:55:07 INFO DistriOptimizer$: [Epoch 14 13824/14923][Iteration 278][Wall Clock 634.620078722s] Trained 768 records in 1.715134668 seconds. Throughput is 447.77826 records/second. Loss is 2.980821. 
19/02/11 21:55:08 INFO DistriOptimizer$: [Epoch 14 14592/14923][Iteration 279][Wall Clock 636.389390068s] Trained 768 records in 1.769311346 seconds. Throughput is 434.0672 records/second. Loss is 2.9798176. 
19/02/11 21:55:10 INFO DistriOptimizer$: [Epoch 14 15360/14923][Iteration 280][Wall Clock 638.16818019s] Trained 768 records in 1.778790122 seconds. Throughput is 431.75415 records/second. Loss is 2.9808552. 
19/02/11 21:55:10 INFO DistriOptimizer$: [Epoch 14 15360/14923][Iteration 280][Wall Clock 638.16818019s] Epoch finished. Wall clock time is 646326.482571 ms
19/02/11 21:55:10 INFO DistriOptimizer$: [Epoch 14 15360/14923][Iteration 280][Wall Clock 638.16818019s] Validate model...
19/02/11 21:55:17 INFO DistriOptimizer$: [Epoch 14 15360/14923][Iteration 280][Wall Clock 638.16818019s] Top1Accuracy is Accuracy(correct: 224, count: 3905, accuracy: 0.05736235595390525)
19/02/11 21:55:19 INFO DistriOptimizer$: [Epoch 15 768/14923][Iteration 281][Wall Clock 648.14874026s] Trained 768 records in 1.822257689 seconds. Throughput is 421.45523 records/second. Loss is 2.984185. 
19/02/11 21:55:21 INFO DistriOptimizer$: [Epoch 15 1536/14923][Iteration 282][Wall Clock 649.942929731s] Trained 768 records in 1.794189471 seconds. Throughput is 428.04843 records/second. Loss is 2.9798527. 
19/02/11 21:55:22 INFO DistriOptimizer$: [Epoch 15 2304/14923][Iteration 283][Wall Clock 651.61417292s] Trained 768 records in 1.671243189 seconds. Throughput is 459.53815 records/second. Loss is 2.974022. 
19/02/11 21:55:24 INFO DistriOptimizer$: [Epoch 15 3072/14923][Iteration 284][Wall Clock 653.355413848s] Trained 768 records in 1.741240928 seconds. Throughput is 441.06473 records/second. Loss is 2.97656. 
19/02/11 21:55:26 INFO DistriOptimizer$: [Epoch 15 3840/14923][Iteration 285][Wall Clock 655.045011572s] Trained 768 records in 1.689597724 seconds. Throughput is 454.54608 records/second. Loss is 2.9848154. 
19/02/11 21:55:28 INFO DistriOptimizer$: [Epoch 15 4608/14923][Iteration 286][Wall Clock 656.788181856s] Trained 768 records in 1.743170284 seconds. Throughput is 440.5766 records/second. Loss is 2.9894907. 
19/02/11 21:55:29 INFO DistriOptimizer$: [Epoch 15 5376/14923][Iteration 287][Wall Clock 658.480493859s] Trained 768 records in 1.692312003 seconds. Throughput is 453.81702 records/second. Loss is 2.9772322. 
19/02/11 21:55:31 INFO DistriOptimizer$: [Epoch 15 6144/14923][Iteration 288][Wall Clock 660.166880089s] Trained 768 records in 1.68638623 seconds. Throughput is 455.41168 records/second. Loss is 2.984101. 
19/02/11 21:55:33 INFO DistriOptimizer$: [Epoch 15 6912/14923][Iteration 289][Wall Clock 661.85638721s] Trained 768 records in 1.689507121 seconds. Throughput is 454.57043 records/second. Loss is 2.982938. 
19/02/11 21:55:34 INFO DistriOptimizer$: [Epoch 15 7680/14923][Iteration 290][Wall Clock 663.620091493s] Trained 768 records in 1.763704283 seconds. Throughput is 435.44714 records/second. Loss is 2.97358. 
19/02/11 21:55:36 INFO DistriOptimizer$: [Epoch 15 8448/14923][Iteration 291][Wall Clock 665.260683388s] Trained 768 records in 1.640591895 seconds. Throughput is 468.12375 records/second. Loss is 2.966507. 
19/02/11 21:55:38 INFO DistriOptimizer$: [Epoch 15 9216/14923][Iteration 292][Wall Clock 666.960838219s] Trained 768 records in 1.700154831 seconds. Throughput is 451.72357 records/second. Loss is 2.9650447. 
19/02/11 21:55:39 INFO DistriOptimizer$: [Epoch 15 9984/14923][Iteration 293][Wall Clock 668.672705791s] Trained 768 records in 1.711867572 seconds. Throughput is 448.63284 records/second. Loss is 2.9792418. 
19/02/11 21:55:41 INFO DistriOptimizer$: [Epoch 15 10752/14923][Iteration 294][Wall Clock 670.321723432s] Trained 768 records in 1.649017641 seconds. Throughput is 465.7318 records/second. Loss is 2.9735532. 
19/02/11 21:55:43 INFO DistriOptimizer$: [Epoch 15 11520/14923][Iteration 295][Wall Clock 672.04684509s] Trained 768 records in 1.725121658 seconds. Throughput is 445.186 records/second. Loss is 2.9840238. 
19/02/11 21:55:45 INFO DistriOptimizer$: [Epoch 15 12288/14923][Iteration 296][Wall Clock 673.80154841s] Trained 768 records in 1.75470332 seconds. Throughput is 437.68085 records/second. Loss is 2.9751852. 
19/02/11 21:55:46 INFO DistriOptimizer$: [Epoch 15 13056/14923][Iteration 297][Wall Clock 675.43471919s] Trained 768 records in 1.63317078 seconds. Throughput is 470.2509 records/second. Loss is 2.9777195. 
19/02/11 21:55:48 INFO DistriOptimizer$: [Epoch 15 13824/14923][Iteration 298][Wall Clock 677.025826644s] Trained 768 records in 1.591107454 seconds. Throughput is 482.68268 records/second. Loss is 2.9705133. 
19/02/11 21:55:50 INFO DistriOptimizer$: [Epoch 15 14592/14923][Iteration 299][Wall Clock 678.736325676s] Trained 768 records in 1.710499032 seconds. Throughput is 448.99176 records/second. Loss is 2.9635608. 
19/02/11 21:55:51 INFO DistriOptimizer$: [Epoch 15 15360/14923][Iteration 300][Wall Clock 680.38851991s] Trained 768 records in 1.652194234 seconds. Throughput is 464.8364 records/second. Loss is 2.972312. 
19/02/11 21:55:51 INFO DistriOptimizer$: [Epoch 15 15360/14923][Iteration 300][Wall Clock 680.38851991s] Epoch finished. Wall clock time is 687604.905848 ms
19/02/11 21:55:51 INFO DistriOptimizer$: [Epoch 15 15360/14923][Iteration 300][Wall Clock 680.38851991s] Validate model...
19/02/11 21:55:58 INFO DistriOptimizer$: [Epoch 15 15360/14923][Iteration 300][Wall Clock 680.38851991s] Top1Accuracy is Accuracy(correct: 213, count: 3905, accuracy: 0.05454545454545454)
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.disk.diskSpaceUsed_MB, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.maxMem_MB, value=102145
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.maxOffHeapMem_MB, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.maxOnHeapMem_MB, value=102145
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.memUsed_MB, value=10153
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.offHeapMemUsed_MB, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.onHeapMemUsed_MB, value=10153
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.remainingMem_MB, value=91991
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.remainingOffHeapMem_MB, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.BlockManager.memory.remainingOnHeapMem_MB, value=91991
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.DAGScheduler.job.activeJobs, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.DAGScheduler.job.allJobs, value=640
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.DAGScheduler.stage.failedStages, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.DAGScheduler.stage.runningStages, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.DAGScheduler.stage.waitingStages, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.PS-MarkSweep.count, value=4
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.PS-MarkSweep.time, value=363
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.PS-Scavenge.count, value=14
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.PS-Scavenge.time, value=437
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.heap.committed, value=4083154944
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.heap.init, value=924844032
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.heap.max, value=28633464832
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.heap.usage, value=0.052959056715529244
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.heap.used, value=1516401288
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.non-heap.committed, value=131137536
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.non-heap.init, value=2555904
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.non-heap.max, value=-1
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.non-heap.usage, value=-1.29285344E8
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.non-heap.used, value=129285408
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Code-Cache.committed, value=38338560
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Code-Cache.init, value=2555904
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Code-Cache.max, value=251658240
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Code-Cache.usage, value=0.15116780598958332
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Code-Cache.used, value=38042624
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Compressed-Class-Space.committed, value=11010048
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Compressed-Class-Space.init, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Compressed-Class-Space.max, value=1073741824
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Compressed-Class-Space.usage, value=0.009983859956264496
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Compressed-Class-Space.used, value=10720088
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Metaspace.committed, value=81788928
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Metaspace.init, value=0
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Metaspace.max, value=-1
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Metaspace.usage, value=0.9845225994403546
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.Metaspace.used, value=80523048
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Eden-Space.committed, value=2864709632
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Eden-Space.init, value=231735296
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Eden-Space.max, value=10462691328
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Eden-Space.usage, value=0.10167260857186591
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Eden-Space.used, value=1063769120
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Old-Gen.committed, value=1132986368
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Old-Gen.init, value=616562688
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Old-Gen.max, value=21474836480
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Old-Gen.usage, value=0.019513317197561265
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Old-Gen.used, value=419045296
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Survivor-Space.committed, value=85458944
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Survivor-Space.init, value=38273024
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Survivor-Space.max, value=85458944
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Survivor-Space.usage, value=0.3930176342923217
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.pools.PS-Survivor-Space.used, value=33586872
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.total.committed, value=4214292480
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.total.init, value=927399936
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.total.max, value=28633464831
19/02/11 21:56:00 INFO metrics: type=GAUGE, name=application_1549761699712_0010.driver.jvm.total.used, value=1645870696
19/02/11 21:56:00 INFO metrics: type=COUNTER, name=application_1549761699712_0010.driver.HiveExternalCatalog.fileCacheHits, count=0
19/02/11 21:56:00 INFO metrics: type=COUNTER, name=application_1549761699712_0010.driver.HiveExternalCatalog.filesDiscovered, count=0
19/02/11 21:56:00 INFO metrics: type=COUNTER, name=application_1549761699712_0010.driver.HiveExternalCatalog.hiveClientCalls, count=0
19/02/11 21:56:00 INFO metrics: type=COUNTER, name=application_1549761699712_0010.driver.HiveExternalCatalog.parallelListingJobCount, count=0
19/02/11 21:56:00 INFO metrics: type=COUNTER, name=application_1549761699712_0010.driver.HiveExternalCatalog.partitionsFetched, count=0
19/02/11 21:56:00 INFO metrics: type=HISTOGRAM, name=application_1549761699712_0010.driver.CodeGenerator.compilationTime, count=0, min=0, max=0, mean=0.0, stddev=0.0, median=0.0, p75=0.0, p95=0.0, p98=0.0, p99=0.0, p999=0.0
19/02/11 21:56:00 INFO metrics: type=HISTOGRAM, name=application_1549761699712_0010.driver.CodeGenerator.generatedClassSize, count=0, min=0, max=0, mean=0.0, stddev=0.0, median=0.0, p75=0.0, p95=0.0, p98=0.0, p99=0.0, p999=0.0
19/02/11 21:56:00 INFO metrics: type=HISTOGRAM, name=application_1549761699712_0010.driver.CodeGenerator.generatedMethodSize, count=0, min=0, max=0, mean=0.0, stddev=0.0, median=0.0, p75=0.0, p95=0.0, p98=0.0, p99=0.0, p999=0.0
19/02/11 21:56:00 INFO metrics: type=HISTOGRAM, name=application_1549761699712_0010.driver.CodeGenerator.sourceCodeSize, count=0, min=0, max=0, mean=0.0, stddev=0.0, median=0.0, p75=0.0, p95=0.0, p98=0.0, p99=0.0, p999=0.0
19/02/11 21:56:00 INFO metrics: type=TIMER, name=application_1549761699712_0010.driver.DAGScheduler.messageProcessingTime, count=10760, min=9.999999999999999E-5, max=9.895289, mean=0.39610863933167473, stddev=1.1862531925728736, median=0.052801999999999995, p75=0.153805, p95=3.748509, p98=4.677436, p99=5.120849, p999=8.462446, mean_rate=9.381494965500876, m1=16.47311897813088, m5=14.75906615830696, m15=8.503909252857786, rate_unit=events/second, duration_unit=milliseconds

[EDIT] I posted the thread's model as Text Classifier, however I'm mentioning Sentiment Analysis. Has anyone ever had this low accuracy on LSTM for Text Classifier application? Thanks!

[wzhongyuan]

which example were you running ?

[Bfzanchetta]

I'm running Sentiment Analysis example on BigDL. I'm using 2 head nodes and 8 workers. Each node contains 16 vCPUs and 120GB of RAM. They are all designed into a Spark2 cluster with Hadoop and YARN.

Interesting points: I was having the same issue on the example Text Classifier for LSTM, where after 15 or 20 epochs it kept ending with <5% accuracy rates. However, I missed a passage on BigDL's wiki that said that training distributed LSTM networks demands higher number of epochs to achieve the same accuracy than a regular CNN model at epoch 15.

I will test this in Sentiment Analysis.

[jason-dai]

For sentiment analysis, you may take a look at https://github.com/intel-analytics/analytics-zoo/tree/master/apps/sentiment-analysis

[YongyiZhou]

Hello @Bfzanchetta, I'm now trying to build a textclassifier model as well. But I met RPC lost while training, but I only used a 9G training set, with a 300G*4 cluster.

Could you show me your "build model" and "optimizer" code so I can figure out weather it's spark config problem or my app problem?

Thanks!