Fetching contributors…
Cannot retrieve contributors at this time
23 lines (17 sloc) 710 Bytes

Benchmarks for DSSTNE

We ran the benchmark of DSSTNE using the Movielens data set. Training Parameter Specific

  • 27278 input/output dimensions
  • 3 Hidden Sigmoid layers with 1024 each
  • RMSProp learner
  • 256 Batchsize

Time taken to run one epoch is considered for performance comparison


Use the Config at config.json. Follow the example and change the training command

train -i -o -d gl -c config.json -b 256 -e 20 -n

TensorFlow -u 1024 -b 256 -i 1082 -v54 --vocab_size 27278 -l 3 -f /input/data/ml20m-all.remotcc