feat(regression): add l1 metric for regression

docs/api.md

@@ -669,7 +669,7 @@ No parameters
 Parameter         | Type   | Optional | Default | Description
 ---------         | ----   | -------- | ------- | -----------
 best              | int    | yes      | 1       | Number of top predictions returned by data URI (supervised)
-measure           | array  | yes      | empty   | Output measures requested, from `acc`: accuracy, `acc-k`: top-k accuracy, replace k with number (e.g. `acc-5`), `f1`: f1, precision and recall, `mcll`: multi-class log loss, `auc`: area under the curve, `cmdiag`: diagonal of confusion matrix (requires `f1`), `cmfull`: full confusion matrix (requires `f1`), `mcc`: Matthews correlation coefficient, `eucll`: euclidean distance (e.g. for regression tasks), `kl`: KL_divergence, `js`: JS divergence, `was`: Wasserstein, `ks`: Kolmogorov Smirnov, `dc`: distance correlation, `r2`: R2, `deltas`: delta scores, 'raw': ouput raw results, in case of predict call, this requires a special deploy.prototxt that is a test network (to have ground truth)
+measure           | array  | yes      | empty   | Output measures requested, from `acc`: accuracy, `acc-k`: top-k accuracy, replace k with number (e.g. `acc-5`), `f1`: f1, precision and recall, `mcll`: multi-class log loss, `auc`: area under the curve, `cmdiag`: diagonal of confusion matrix (requires `f1`), `cmfull`: full confusion matrix (requires `f1`), `mcc`: Matthews correlation coefficient, `eucll`: euclidean distance (e.g. for regression tasks),`l1`: l1 distance (e.g. for regression tasks), `kl`: KL_divergence, `js`: JS divergence, `was`: Wasserstein, `ks`: Kolmogorov Smirnov, `dc`: distance correlation, `r2`: R2, `deltas`: delta scores, 'raw': ouput raw results, in case of predict call, this requires a special deploy.prototxt that is a test network (to have ground truth)
 target_repository | string | yes      | empty   | target directory to which to copy the best model files once training has completed
 
 #### Machine learning libraries

src/supervisedoutputconnector.h

@@ -780,7 +780,10 @@ namespace dd
           bool beucll = false;
           float beucll_thres = -1;
           find_presence_and_thres("eucll", measures, beucll, beucll_thres);
-          bool compute_all_meucll = beucll && !autoencoder;
+          bool bl1 = (std::find(measures.begin(), measures.end(), "l1")
+                      != measures.end());
+          bool compute_all_distl = (beucll || bl1) && !autoencoder;
+
           bool bmcc = (std::find(measures.begin(), measures.end(), "mcc")
                        != measures.end());
           bool baccv = false;
@@ -1074,16 +1077,16 @@ namespace dd
               double meucll;
               std::vector<double> all_meucll;
               std::tie(meucll, all_meucll)
-                  = eucll(ad_res, -1, compute_all_meucll);
+                  = distl(ad_res, -1, compute_all_distl, false);
               meas_out.add("eucll", meucll);
-              if (all_meucll.size() > 1 && compute_all_meucll)
+              if (all_meucll.size() > 1 && compute_all_distl)
                 for (unsigned int i = 0; i < all_meucll.size(); ++i)
                   meas_out.add("eucll_" + std::to_string(i), all_meucll[i]);
 
               if (beucll_thres > 0)
                 {
                   std::tuple<double, std::vector<double>> tmeucll_thres
-                      = eucll(ad_res, beucll_thres, compute_all_meucll);
+                      = distl(ad_res, beucll_thres, compute_all_distl, false);
                   double meucll_thres = std::get<0>(tmeucll_thres);
                   std::string b = "eucll_no_" + std::to_string(beucll_thres);
                   meas_out.add(b, meucll_thres);
@@ -1098,6 +1101,16 @@ namespace dd
                       }
                 }
             }
+          if (bl1)
+            {
+              double ml1;
+              std::vector<double> all_ml1;
+              std::tie(ml1, all_ml1)
+                  = distl(ad_res, -1, compute_all_distl, true);
+              meas_out.add("l1", ml1);
+              for (unsigned int i = 0; i < all_ml1.size(); ++i)
+                meas_out.add("l1_" + std::to_string(i), all_ml1[i]);
+            }
           if (bmcc)
             {
               double mmcc = mcc(ad_res);
@@ -2574,15 +2587,16 @@ namespace dd
     }
 
     static std::tuple<double, std::vector<double>>
-    eucll(const APIData &ad, float thres, bool compute_all_meucll)
+    distl(const APIData &ad, float thres, bool compute_all_distl,
+          bool l1 = false)
     {
       double eucl = 0.0;
       unsigned int psize = ad.getobj(std::to_string(0))
                                .get("pred")
                                .get<std::vector<double>>()
                                .size();
       std::vector<double> all_eucl;
-      if (compute_all_meucll)
+      if (compute_all_distl)
         all_eucl.resize(psize, 0.0);
       int batch_size = ad.get("batch_size").get<int>();
       bool has_ignore = ad.has("ignore_label");
@@ -2612,22 +2626,39 @@ namespace dd
                 {
                   if (diff >= thres)
                     {
-                      leucl += diff * diff;
-                      if (compute_all_meucll)
+                      if (l1)
+                        eucl += diff;
+                      else
+                        leucl += diff * diff;
+                      if (compute_all_distl)
                         all_eucl[j] += diff;
                     }
                 }
               else
                 {
-                  leucl += diff * diff;
-                  if (compute_all_meucll)
-                    all_eucl[j] += diff;
+                  if (l1)
+                    eucl += diff;
+                  else
+                    leucl += diff * diff;
+                  if (compute_all_distl)
+                    {
+                      if (l1)
+                        all_eucl[j] += diff;
+                      else
+                        all_eucl[j] += diff * diff;
+                    }
                 }
             }
-          eucl += sqrt(leucl);
+          if (!l1)
+            {
+              eucl += sqrt(leucl);
+              if (compute_all_distl)
+                for (size_t j = 0; j < target.size(); ++j)
+                  all_eucl[j] = sqrt(all_eucl[j]);
+            }
         }
 
-      if (compute_all_meucll)
+      if (compute_all_distl)
         for (unsigned int i = 0; i < all_eucl.size(); ++i)
           all_eucl[i] /= static_cast<double>(batch_size);
 

tests/ut-torchapi.cc

@@ -1996,6 +1996,68 @@ TEST(torchapi, service_train_images_split_regression_2dims_db_false)
   fileops::remove_dir(resnet50_train_repo + "test_0.lmdb");
 }
 
+TEST(torchapi, service_train_images_split_regression_2dims_db_false_l1)
+{
+  setenv("CUBLAS_WORKSPACE_CONFIG", ":4096:8", true);
+  torch::manual_seed(torch_seed);
+  at::globalContext().setDeterministicCuDNN(true);
+
+  // Create service
+  JsonAPI japi;
+  std::string sname = "imgserv";
+  std::string jstr
+      = "{\"mllib\":\"torch\",\"description\":\"image\",\"type\":"
+        "\"supervised\",\"model\":{\"repository\":\""
+        + resnet50_train_repo
+        + "\"},\"parameters\":{\"input\":{\"connector\":\"image\","
+          "\"width\":224,\"height\":224,\"db\":false},\"mllib\":{\"ntargets\":"
+          "2,\"finetuning\":true,\"regression\":true,\"gpu\":true,\"loss\":"
+          "\"L2\"}}}";
+  std::string joutstr = japi.jrender(japi.service_create(sname, jstr));
+  ASSERT_EQ(created_str, joutstr);
+
+  // Train
+  std::string jtrainstr
+      = "{\"service\":\"imgserv\",\"async\":false,\"parameters\":{"
+        "\"mllib\":{\"solver\":{\"iterations\":"
+        + iterations_resnet50 + ",\"base_lr\":" + torch_lr
+        + ",\"iter_size\":4,\"solver_"
+          "type\":\"ADAM\",\"test_interval\":200},\"net\":{\"batch_size\":4, "
+          "\"test_batch_size\":3},"
+          "\"resume\":false},"
+          "\"input\":{\"seed\":12345,\"db\":false,\"shuffle\":true,\"test_"
+          "split\":0.1},"
+          "\"output\":{\"measure\":[\"l1\"]}},\"data\":[\""
+        + resnet50_train_data_reg2 + "\"]}";
+  joutstr = japi.jrender(japi.service_train(jtrainstr));
+  JDoc jd;
+  std::cout << "joutstr=" << joutstr << std::endl;
+  jd.Parse<rapidjson::kParseNanAndInfFlag>(joutstr.c_str());
+  ASSERT_TRUE(!jd.HasParseError());
+  ASSERT_EQ(201, jd["status"]["code"]);
+
+  ASSERT_TRUE(jd["body"]["measure"]["iteration"] == 200) << "iterations";
+  ASSERT_TRUE(jd["body"]["measure"]["l1"].GetDouble() <= 15.0) << "l1";
+
+  std::unordered_set<std::string> lfiles;
+  fileops::list_directory(resnet50_train_repo, true, false, false, lfiles);
+  for (std::string ff : lfiles)
+    {
+      if (ff.find("checkpoint") != std::string::npos
+          || ff.find("solver") != std::string::npos)
+        remove(ff.c_str());
+    }
+  ASSERT_TRUE(!fileops::file_exists(resnet50_train_repo + "checkpoint-"
+                                    + iterations_resnet50 + ".ptw"));
+  ASSERT_TRUE(!fileops::file_exists(resnet50_train_repo + "checkpoint-"
+                                    + iterations_resnet50 + ".pt"));
+
+  fileops::clear_directory(resnet50_train_repo + "train.lmdb");
+  fileops::clear_directory(resnet50_train_repo + "test_0.lmdb");
+  fileops::remove_dir(resnet50_train_repo + "train.lmdb");
+  fileops::remove_dir(resnet50_train_repo + "test_0.lmdb");
+}
+
 TEST(torchapi, service_train_object_detection_fasterrcnn)
 {
   setenv("CUBLAS_WORKSPACE_CONFIG", ":4096:8", true);