Merge branch 'master' of https://github.com/ECP-Candle/Supervisor

Author: brettinanl

workflows/nt3_mlrMBO/R/mlrMBO3.R

@@ -59,9 +59,9 @@ main_function <- function(max.budget = 110, max.iterations = 10, design.size=10,
 
 
   surr.rf = makeLearner("regr.randomForest", predict.type = "se")
-  ctrl = makeMBOControl(n.objectives = 1, propose.points = 20, 
+  ctrl = makeMBOControl(n.objectives = 1, propose.points = min(20, propose.points), 
        	  			       impute.y.fun = function(x, y, opt.path, ...) .Machine$integer.max * 0.1 )
-  ctrl = setMBOControlInfill(ctrl, crit=makeMBOInfillCritCB(), interleave.random.points=propose.points-20)
+  ctrl = setMBOControlInfill(ctrl, crit=makeMBOInfillCritCB(), interleave.random.points=max(0,propose.points-20))
   ctrl = setMBOControlMultiPoint(ctrl, method = "cb")
   ctrl = setMBOControlTermination(ctrl, max.evals = max.budget)
   ctrl = setMBOControlTermination(ctrl, iters = max.iterations)

workflows/nt3_mlrMBO/data/parameter_set3.R

@@ -6,7 +6,7 @@ param.set <- makeParamSet(
   makeIntegerParam("epochs", lower = 5, upper = 500),
   makeDiscreteParam("activation", values = c("softmax", "elu", "softplus", "softsign", "relu", "tanh", "sigmoid", "hard_sigmoid", "linear")),
   makeDiscreteParam("dense", values = c("500 100 50", "1000 500 100 50", "2000 1000 500 100 50", "2000 1000 1000 500 100 50", "2000 1000 1000 1000 500 100 50")),
-  makeDiscreteParam("optimizer", values = c("adam", "sgd", "rmsprop", "adagrad", "adadelta","adamax","nadam")),
+  makeDiscreteParam("optimizer", values = c("adam", "sgd", "rmsprop", "adagrad", "adadelta")),
   makeNumericParam("drop", lower = 0, upper = 0.9),
   makeNumericParam("learning_rate", lower = 0.00001, upper = 0.1),
   makeDiscreteParam("conv", values = c("50 50 50 50 50 1", "25 25 25 25 25 1", "64 32 16 32 64 1", "100 100 100 100 100 1", "32 20 16 32 10 1"))

workflows/nt3_mlrMBO/swift/ai_workflow3.swift

@@ -68,17 +68,23 @@ app (file out, file err) run_log_end(file shfile)
   o = propagate();
 }
 
+(string obj_result) get_results(string result_file) {
+  if (file_exists(result_file)) {
+    file line = input(result_file);
+    obj_result = trim(read(line));
+  } else {
+    obj_result = "NaN";
+  }
+}
+
 (string obj_result) obj(string params, string iter_indiv_id) {
   string outdir = "%s/run_%s" % (turbine_output, iter_indiv_id);
   file out <"%s/out.txt" % outdir>;
   file err <"%s/err.txt" % outdir>;
 
-  //printf("%s", params);
-  //string fname = "%s/params.json" % outdir =>
-  //file params_file <fname> = write(params) =>
   (out,err) = run_model(model_script, params, outdir, iter_indiv_id) =>
-  file line = input("%s/result.txt" % outdir) =>
-  obj_result = trim(read(line));
+  string result_file = "%s/result.txt" % outdir;
+  obj_result = get_results(result_file);
   printf(obj_result);
 }
 

workflows/p2b1_mlrMBO/R/mlrMBO3.R

@@ -57,10 +57,15 @@ simple.obj.fun = function(x){}
 
 main_function <- function(max.budget = 110, max.iterations = 10, design.size=10, propose.points=10){
 
+
   surr.rf = makeLearner("regr.randomForest", predict.type = "se")
-  ctrl = makeMBOControl(n.objectives = 1, propose.points = propose.points)
+  ctrl = makeMBOControl(n.objectives = 1, propose.points = min(20, propose.points), 
+       	   			       	         impute.y.fun = function(x, y, opt.path, ...) .Machine$integer.max * 0.1 )
+  ctrl = setMBOControlInfill(ctrl, crit=makeMBOInfillCritCB(), interleave.random.points=max(0,propose.points-20))
+  ctrl = setMBOControlMultiPoint(ctrl, method = "cb")
   ctrl = setMBOControlTermination(ctrl, max.evals = max.budget)
   ctrl = setMBOControlTermination(ctrl, iters = max.iterations)
+
   # ctrl = setMBOControlInfill(ctrl, crit =makeMBOInfillCritCB(), opt.focussearch.points = 500)
   design = generateDesign(n = design.size, par.set = getParamSet(obj.fun))
   configureMlr(show.info = FALSE, show.learner.output = FALSE, on.learner.warning = "quiet")

workflows/p2b1_mlrMBO/data/parameter_set3.R

@@ -2,6 +2,21 @@
 # the parameter names should match names of the arguments expected by
 # the benchmark
 param.set <- makeParamSet(
-  makeDiscreteParam("batch_size", values = c(32, 64, 128)),
-  makeIntegerParam("epochs", lower = 1, upper = 3)
+  makeDiscreteParam("batch_size", values = c(16, 32, 64, 128, 256, 512)),
+  makeIntegerParam("epochs", lower = 1, upper = 2),
+  makeDiscreteParam("activation", values = c("softmax","elu","softplus","softsign", 
+  				  	   	"relu", "tanh","sigmoid","hard_sigmoid",
+						"linear")),
+  makeDiscreteParam("optimizer", values = c("adam", "sgd", "rmsprop","adagrad",
+  				 	  	    "adadelta")),
+  makeNumericParam("learning_rate", lower = 0.001, upper = 0.2),
+  makeNumericParam("noise", lower = 0.001, upper = 0.5)
+  #makeDiscreteParam("num_hidden", values = c("512 256 128 64 32 16", "512 64")),
+  #makeDiscreteParam("molecular_num_hidden", values = c("54 12", "32 16 8"))
+  ## DEBUG PARAMETERS: DON'T USE THESE IN PRODUCTION RUN
+  ## makeDiscreteParam("conv", values = c("32 20 16 32 10 1"))
 )
+
+
+
+

workflows/p2b1_mlrMBO/scripts/theta_run_model.sh

@@ -73,6 +73,7 @@ MODEL_CMD="python ${arg_array[@]}"
 set +e
 echo $MODEL_CMD
 $TIMEOUT_CMD python "${arg_array[@]}"
+sleep 60
 
 RES=$?
 if [ "$RES" -ne 0 ]; then

workflows/p2b1_mlrMBO/swift/ai_workflow3.swift

@@ -68,19 +68,30 @@ app (file out, file err) run_log_end(file shfile)
   o = propagate();
 }
 
+
+(string obj_result) get_results(string result_file) {
+  if (file_exists(result_file)) {
+    file line = input(result_file);
+    obj_result = trim(read(line));
+  } else {
+    obj_result = "NaN";
+  }
+}
+
 (string obj_result) obj(string params, string iter_indiv_id) {
   string outdir = "%s/run_%s" % (turbine_output, iter_indiv_id);
   file out <"%s/out.txt" % outdir>;
   file err <"%s/err.txt" % outdir>;
 
-  //string fname = "%s/params.json" % outdir =>
-  //file params_file <fname> = write(params) =>
   (out,err) = run_model(model_script, params, outdir, iter_indiv_id) =>
-  file line = input("%s/result.txt" % outdir) =>
-  obj_result = trim(read(line));
+  string result_file = "%s/result.txt" % outdir;
+  obj_result = get_results(result_file);
   printf(obj_result);
 }
 
+
+
+
 (void v) loop(location ME, int ME_rank) {
 
     for (boolean b = true, int i = 1;

workflows/p3b1_mlrMBO/R/mlrMBO3.R

@@ -58,9 +58,13 @@ simple.obj.fun = function(x){}
 main_function <- function(max.budget = 110, max.iterations = 10, design.size=10, propose.points=10){
 
   surr.rf = makeLearner("regr.randomForest", predict.type = "se")
-  ctrl = makeMBOControl(n.objectives = 1, propose.points = propose.points)
+  ctrl = makeMBOControl(n.objectives = 1, propose.points = min(20, propose.points), 
+       	   			       	         impute.y.fun = function(x, y, opt.path, ...) .Machine$integer.max * 0.1 )
+  ctrl = setMBOControlInfill(ctrl, crit=makeMBOInfillCritCB(), interleave.random.points=max(0,propose.points-20))
+  ctrl = setMBOControlMultiPoint(ctrl, method = "cb")
   ctrl = setMBOControlTermination(ctrl, max.evals = max.budget)
   ctrl = setMBOControlTermination(ctrl, iters = max.iterations)
+
   # ctrl = setMBOControlInfill(ctrl, crit =makeMBOInfillCritCB(), opt.focussearch.points = 500)
   design = generateDesign(n = design.size, par.set = getParamSet(obj.fun))
   configureMlr(show.info = FALSE, show.learner.output = FALSE, on.learner.warning = "quiet")

workflows/p3b1_mlrMBO/data/parameter_set3.R

@@ -1,6 +1,13 @@
 # see https://cran.r-project.org/web/packages/ParamHelpers/ParamHelpers.pdfmakeNum
-# the parameter names should match names of the arguments expected by
-# the benchmark
+# the parameter names should match names of the arguments expected by the benchmark
+
 param.set <- makeParamSet(
-  makeIntegerParam("epochs", lower = 1, upper = 3)
+  makeDiscreteParam("batch_size", values = c(16, 32, 64, 128, 256, 512)),
+  makeIntegerParam("epochs", lower = 5, upper = 500),
+  makeDiscreteParam("activation", values = c("softmax", "elu", "softplus", "softsign", "relu", "tanh", "sigmoid", "hard_sigmoid", "linear")),
+  makeDiscreteParam("optimizer", values = c("adam", "sgd", "rmsprop", "adagrad", "adadelta")),
+  makeNumericParam("dropout", lower = 0, upper = 0.9),
+  makeNumericParam("learning_rate", lower = 0.00001, upper = 0.1)
+  ## DEBUG PARAMETERS: DON'T USE THESE IN PRODUCTION RUN
+  ## makeDiscreteParam("conv", values = c("32 20 16 32 10 1"))
 )

workflows/p3b1_mlrMBO/swift/ai_workflow3.swift

@@ -67,19 +67,28 @@ app (file out, file err) run_log_end(file shfile)
   o = propagate();
 }
 
+
+(string obj_result) get_results(string result_file) {
+  if (file_exists(result_file)) {
+    file line = input(result_file);
+    obj_result = trim(read(line));
+  } else {
+    obj_result = "NaN";
+  }
+}
+
 (string obj_result) obj(string params, string iter_indiv_id) {
   string outdir = "%s/run_%s" % (turbine_output, iter_indiv_id);
   file out <"%s/out.txt" % outdir>;
   file err <"%s/err.txt" % outdir>;
 
-  //string fname = "%s/params.json" % outdir =>
-  //file params_file <fname> = write(params) =>
   (out,err) = run_model(model_script, params, outdir, iter_indiv_id) =>
-  file line = input("%s/result.txt" % outdir) =>
-  obj_result = trim(read(line));
+  string result_file = "%s/result.txt" % outdir;
+  obj_result = get_results(result_file);
   printf(obj_result);
 }
 
+
 (void v) loop(location ME, int ME_rank) {
 
     for (boolean b = true, int i = 1;