make data reader single threaded

constants_crunchy.py

@@ -8,7 +8,7 @@ def mkdir_if_not_exists(dir_name):
 experiment_name = 'QA_explicit_dot_joint_training_pretrained_same_lr'
 #experiment_name = 'object_attribute_classifier_large_images'
 # Global output directory (all subexperiments will be saved here)
-global_output_dir = '/home/tanmay/Code/GenVQA/Exp_Results/VQA'
+global_output_dir = '/data/tanmay/GenVQA_Exp_Results'
 
 global_experiment_dir = os.path.join(
     global_output_dir,

constants_vision_gpu_2.py

@@ -8,7 +8,7 @@ def mkdir_if_not_exists(dir_name):
 experiment_name = 'QA_explicit_dot_pretrained_same_lr'
 #experiment_name = 'object_attribute_classifier_large_images'
 # Global output directory (all subexperiments will be saved here)
-global_output_dir = '/home/nfs/tgupta6/projects/GenVQA/Exp_Results/VQA'
+global_output_dir = '/data/tanmay/GenVQA_Exp_Results'
 
 global_experiment_dir = os.path.join(
     global_output_dir,
@@ -164,7 +164,7 @@ answer_obj_atr_loss_wt = 0.0
 answer_regularization_coeff = 1e-5
 answer_queue_size = 500
 answer_embedding_dim = 600
-answer_lr = 1e-3
+answer_lr = 1e-4
 answer_log_every_n_iter = 500
 answer_output_dir = os.path.join(
     global_experiment_dir,
@@ -184,7 +184,7 @@ answer_model = os.path.join(
 num_regions_with_labels = 100
 
 # Answer fine tune params
-answer_fine_tune_from_iter = 62500
+answer_fine_tune_from_iter = 19500
 answer_fine_tune_from = answer_model + '-' + str(answer_fine_tune_from_iter)
 
 # Answer eval params

data/cropped_regions_cached_features.py

@@ -87,20 +87,22 @@ class data():
             print 'Error in thread {}: {}'.format(
                 threading.current_thread().name, str(e))
 
+
     def get_parallel(self, samples):
         batch_list = [None]*len(samples)
         worker_ids = range(len(samples))
         workers = []
         for count, sample in enumerate(samples):
-            worker = threading.Thread(
-                target = self.get_single, 
-                args = (sample, batch_list, worker_ids[count]))
-            worker.setDaemon(True)
-            worker.start()
-            workers.append(worker)
+            self.get_single(sample, batch_list, worker_ids[count])
+        #     worker = threading.Thread(
+        #         target = self.get_single, 
+        #         args = (sample, batch_list, worker_ids[count]))
+        #     worker.setDaemon(True)
+        #     worker.start()
+        #     workers.append(worker)
         
-        for worker in workers:
-            worker.join()
+        # for worker in workers:
+        #     worker.join()
 
         batch_size = len(samples)
         batch = dict()