From e18bb781bc137e540a373e0ce2e324ff8d4d804e Mon Sep 17 00:00:00 2001
From: innerNULL <lyt697@outlook.com>
Date: Sat, 28 Mar 2020 15:35:00 +0800
Subject: [PATCH] Supports finetuning mode.

---
 src/libfm/libfm.cpp | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)

diff --git a/src/libfm/libfm.cpp b/src/libfm/libfm.cpp
index 82b795a..57e01b1 100644
--- a/src/libfm/libfm.cpp
+++ b/src/libfm/libfm.cpp
@@ -100,11 +100,12 @@ int main(int argc, char **argv) {
 
     const std::string param_save_model = cmdline.registerParameter("save_model", "filename for writing the FM model");
     const std::string param_load_model = cmdline.registerParameter("load_model", "filename for reading the FM model");
-
+    const std::string param_finetuning = cmdline.registerParameter("finetuning", "if finetuning mode, 1 for yes, 0 for not, default 0");
     const std::string param_do_sampling  = "do_sampling";
     const std::string param_do_multilevel  = "do_multilevel";
     const std::string param_num_eval_cases  = "num_eval_cases";
 
+
     if (cmdline.hasParameter(param_help) || (argc == 1)) {
       cmdline.print_help();
       return 0;
@@ -118,6 +119,7 @@ int main(int argc, char **argv) {
     if (! cmdline.hasParameter(param_method)) { cmdline.setValue(param_method, "mcmc"); }
     if (! cmdline.hasParameter(param_init_stdev)) { cmdline.setValue(param_init_stdev, "0.1"); }
     if (! cmdline.hasParameter(param_dim)) { cmdline.setValue(param_dim, "1,1,8"); }
+    if (! cmdline.hasParameter(param_finetuning)) { cmdline.setValue(param_finetuning, "0"); }
 
     // Check for invalid flags.
     if (! cmdline.getValue(param_method).compare("mcmc") && cmdline.hasParameter(param_save_model)) {
@@ -132,6 +134,13 @@ int main(int argc, char **argv) {
       return 0;
     }
 
+    if ((!cmdline.hasParameter(param_load_model))
+        && (cmdline.getValue(param_finetuning) == "1")) {
+      std::cout << "WARNING: -finetuning disabled since no loading model." << std::endl;
+      cmdline.setValue(param_finetuning, "0");
+      return 0;
+    }
+
     if (! cmdline.getValue(param_method).compare("als")) { // als is an mcmc without sampling and hyperparameter inference
       cmdline.setValue(param_method, "mcmc");
       if (! cmdline.hasParameter(param_do_sampling)) { cmdline.setValue(param_do_sampling, "0"); }
@@ -280,7 +289,15 @@ int main(int argc, char **argv) {
       ((fm_learn_sgd_element_adapt_reg*)fml)->validation = validation;
 
     } else if (! cmdline.getValue(param_method).compare("mcmc")) {
-      fm.w.init_normal(fm.init_mean, fm.init_stdev);
+      if (cmdline.getValue(param_finetuning) != "1") {
+        std::cout << "WARNING: not finetuning mode. finetuning=" << cmdline.getValue(param_finetuning) << std::endl;
+        fm.w.init_normal(fm.init_mean, fm.init_stdev);
+      } else {
+        std::cout << "WARNING: finetuning mode. finetuning=" << cmdline.getValue(param_finetuning) << std::endl;
+	//std::cout << "DEBUG: debug finetune parameter" << std::endl;
+	//fm.saveModel("./last_model_0");
+	//return 0;
+      }
       fml = new fm_learn_mcmc_simultaneous();
       fml->validation = validation;
       ((fm_learn_mcmc*)fml)->num_iter = cmdline.getValue(param_num_iter, 100);