bool arguments fix

saransh-mehta · saransh-mehta · commit 2a6db66a5a4c · 2020-06-13T16:08:03.000Z
diff --git a/examples/answerability_detection/answerability_detection_msmarco.ipynb b/examples/answerability_detection/answerability_detection_msmarco.ipynb
@@ -141,11 +141,11 @@
     "    --eval_batch_size 16 \\\n",
     "    --grad_accumulation_steps 2 \\\n",
     "    --log_per_updates 250 \\\n",
-    "    --save_per_updates 16000 \\\n",
-    "    --eval_while_train True \\\n",
-    "    --test_while_train True \\\n",
     "    --max_seq_len 324 \\\n",
-    "    --silent True "
+    "    --save_per_updates 16000 \\\n",
+    "    --eval_while_train \\\n",
+    "    --test_while_train \\\n",
+    "    --silent"
    ]
   },
   {
diff --git a/examples/entailment_detection/entailment_snli.ipynb b/examples/entailment_detection/entailment_snli.ipynb
@@ -139,10 +139,10 @@
     "    --eval_batch_size 64 \\\n",
     "    --grad_accumulation_steps 1 \\\n",
     "    --log_per_updates 100 \\\n",
-    "    --eval_while_train True \\\n",
-    "    --test_while_train True \\\n",
     "    --max_seq_len 128 \\\n",
-    "    --silent True "
+    "    --eval_while_train \\\n",
+    "    --test_while_train \\\n",
+    "    --silent"
    ]
   },
   {
diff --git a/examples/intent_ner_fragment/intent_ner_fragment.ipynb b/examples/intent_ner_fragment/intent_ner_fragment.ipynb
@@ -168,7 +168,6 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "\n",
     "!python ../../train.py \\\n",
     "    --data_dir '../../data/bert-base-uncased_prepared_data' \\\n",
     "    --task_file 'tasks_file_snips.yml' \\\n",
@@ -178,10 +177,10 @@
     "    --eval_batch_size 32 \\\n",
     "    --grad_accumulation_steps 2 \\\n",
     "    --log_per_updates 50 \\\n",
-    "    --eval_while_train True \\\n",
-    "    --test_while_train True \\\n",
     "    --max_seq_len 50 \\\n",
-    "    --silent True "
+    "    --eval_while_train \\\n",
+    "    --test_while_train \\\n",
+    "    --silent "
    ]
   },
   {
diff --git a/examples/ner_pos_tagging/ner_pos_tagging_conll.ipynb b/examples/ner_pos_tagging/ner_pos_tagging_conll.ipynb
@@ -58,10 +58,10 @@
     "    --eval_batch_size 32 \\\n",
     "    --grad_accumulation_steps 1 \\\n",
     "    --log_per_updates 50 \\\n",
-    "    --eval_while_train True \\\n",
-    "    --test_while_train True \\\n",
     "    --max_seq_len 50 \\\n",
-    "    --silent True "
+    "    --eval_while_train \\\n",
+    "    --test_while_train \\\n",
+    "    --silent"
    ]
   },
   {
diff --git a/examples/query_correctness/query_correctness.ipynb b/examples/query_correctness/query_correctness.ipynb
@@ -58,10 +58,10 @@
     "    --eval_batch_size 32 \\\n",
     "    --grad_accumulation_steps 1 \\\n",
     "    --log_per_updates 20 \\\n",
-    "    --eval_while_train True \\\n",
-    "    --test_while_train True \\\n",
     "    --max_seq_len 50 \\\n",
-    "    --silent True"
+    "    --eval_while_train \\\n",
+    "    --test_while_train \\\n",
+    "    --silent"
    ]
   },
   {
diff --git a/examples/query_pair_similarity/query_similarity_qqp.ipynb b/examples/query_pair_similarity/query_similarity_qqp.ipynb
@@ -96,9 +96,9 @@
     "    --eval_batch_size 16 \\\n",
     "    --grad_accumulation_steps 2 \\\n",
     "    --log_per_updates 50 \\\n",
-    "    --eval_while_train True \\\n",
-    "    --test_while_train True \\\n",
     "    --max_seq_len 200 \\\n",
+    "    --eval_while_train \\\n",
+    "    --test_while_train \\\n",
     "    --silent True "
    ]
   },
diff --git a/examples/query_type_detection/query_type_detection.ipynb b/examples/query_type_detection/query_type_detection.ipynb
@@ -59,10 +59,10 @@
     "    --eval_batch_size 64 \\\n",
     "    --grad_accumulation_steps 1 \\\n",
     "    --log_per_updates 100 \\\n",
-    "    --eval_while_train True \\\n",
-    "    --test_while_train True \\\n",
     "    --max_seq_len 60 \\\n",
-    "    --silent True"
+    "    --eval_while_train \\\n",
+    "    --test_while_train \\\n",
+    "    --silent"
    ]
   },
   {
diff --git a/models/model.py b/models/model.py
@@ -225,9 +225,8 @@ def update_step(self, batchMetaData, batchData):
         if self.lossClassList[taskId] and (target is not None):
             self.taskLoss = self.lossClassList[taskId](logits, target, attnMasks=modelInputs[2])
             #tensorboard details
-            if self.params['tensorboard']:
-                self.tbTaskId = taskId
-                self.tbTaskLoss = self.taskLoss.item()
+            self.tbTaskId = taskId
+            self.tbTaskLoss = self.taskLoss.item()
         taskLoss = self.taskLoss / self.params['grad_accumulation_steps']
         taskLoss.backward()
         self.accumulatedStep += 1
diff --git a/train.py b/train.py
@@ -29,51 +29,48 @@ def make_arguments(parser):
                         help = 'path to save the model')
     parser.add_argument('--epochs', type = int, required=True,
                         help = 'number of epochs to train')
-    parser.add_argument('--finetune', type = bool, default= False,
-                        help = "If only the shared model is to be loaded with saved pre-trained multi-task model.\
-                            In this case, you can specify your own tasks with task file and use the pre-trained shared model\
-                            to finetune upon.")
-    parser.add_argument('--freeze_shared_model', type = bool, default=False,
+    parser.add_argument('--freeze_shared_model', default=False, action='store_true',
                         help = "True to freeze the loaded pre-trained shared model and only finetune task specific headers")
     parser.add_argument('--train_batch_size', type = int, default=8,
                         help='batch size to use for training')
     parser.add_argument('--eval_batch_size', type = int, default = 32,
                         help = "batch size to use during evaluation")
-    parser.add_argument('--eval_while_train', type = bool, default= True,
-                        help = "if evaluation on dev set is required during training.")
-    parser.add_argument('--test_while_train', type = bool, default = True,
-                        help = "if evaluation on test set is required during training.")
     parser.add_argument('--grad_accumulation_steps', type =int, default = 1,
                         help = "number of steps to accumulate gradients before update")
     parser.add_argument('--num_of_warmup_steps', type=int, default = 0,
                         help = "warm-up value for scheduler")
     parser.add_argument('--grad_clip_value', type = float, default=1.0,
                         help = "gradient clipping value to avoid gradient overflowing" )
-    parser.add_argument('--debug_mode', default = False, type = bool,
-                        help = "record logs for debugging if True")
     parser.add_argument('--log_file', default='multi_task_logs.log', type = str,
                         help = "name of log file to store")
     parser.add_argument('--log_per_updates', default = 10, type = int,
                         help = "number of steps after which to log loss")
-    parser.add_argument('--silent', type = bool, default = True,
-                        help = "Only write logs to file if True")
     parser.add_argument('--seed', default=42, type = int,
                         help = "seed to set for modules")
     parser.add_argument('--max_seq_len', default=128, type =int,
                         help = "max seq length used for model at time of data preparation")
-    parser.add_argument('--tensorboard', default=True, type = bool,
-                        help = "To create tensorboard logs")
     parser.add_argument('--save_per_updates', default = 0, type = int,
                         help = "to keep saving model after this number of updates")
     parser.add_argument('--limit_save', default = 10, type = int,
                         help = "max number recent checkpoints to keep saved")
     parser.add_argument('--load_saved_model', type=str, default=None,
                         help="path to the saved model in case of loading from saved")
-    parser.add_argument('--resume_train', type=bool, default=False, 
-                        help="True for resuming training from a saved model")
+    parser.add_argument('--eval_while_train', default = False, action = 'store_true',
+                        help = "if evaluation on dev set is required during training.")
+    parser.add_argument('--test_while_train', default=False, action = 'store_true',
+                        help = "if evaluation on test set is required during training.")
+    parser.add_argument('--resume_train', default=False, action = 'store_true',
+                        help="Set for resuming training from a saved model")
+    parser.add_argument('--finetune', default= False, action = 'store_true',
+                        help = "If only the shared model is to be loaded with saved pre-trained multi-task model.\
+                            In this case, you can specify your own tasks with task file and use the pre-trained shared model\
+                            to finetune upon.")
+    parser.add_argument('--debug_mode', default = False, action = 'store_true', 
+                        help = "record logs for debugging if True")
+    parser.add_argument('--silent', default = False, action = 'store_true', 
+                        help = "Only write logs to file if True")
     return parser
     
-    
 parser = argparse.ArgumentParser()
 parser = make_arguments(parser)
 args = parser.parse_args()
@@ -193,9 +190,8 @@ def main():
     allParams['gpu'] = torch.cuda.is_available()
     logger.info('task parameters:\n {}'.format(taskParams.taskDetails))
 
-    if args.tensorboard:
-        tensorboard = SummaryWriter(log_dir = os.path.join(logDir, 'tb_logs'))
-        logger.info("Tensorboard writing at {}".format(os.path.join(logDir, 'tb_logs')))
+    tensorboard = SummaryWriter(log_dir = os.path.join(logDir, 'tb_logs'))
+    logger.info("Tensorboard writing at {}".format(os.path.join(logDir, 'tb_logs')))
 
     # making handlers for train
     logger.info("Creating data handlers for training...")
@@ -268,11 +264,11 @@ def main():
                                                                                     taskName,
                                                                                     avgLoss,
                                                                                     model.taskLoss.item()))
-                    if args.tensorboard:
-                        tensorboard.add_scalar('train/avg_loss', avgLoss, global_step= model.globalStep)
-                        tensorboard.add_scalar('train/{}_loss'.format(taskName),
-                                                model.taskLoss.item(),
-                                                global_step=model.globalStep)
+                    
+                    tensorboard.add_scalar('train/avg_loss', avgLoss, global_step= model.globalStep)
+                    tensorboard.add_scalar('train/{}_loss'.format(taskName),
+                                            model.taskLoss.item(),
+                                            global_step=model.globalStep)
                 
                 if args.save_per_updates > 0 and  ( (model.globalStep+1) % args.save_per_updates)==0 and (model.accumulatedStep+1==args.grad_accumulation_steps):
                     savePath = os.path.join(args.out_dir, 'multi_task_model_{}_{}.pt'.format(epoch,
diff --git a/utils/tranform_functions.py b/utils/tranform_functions.py
@@ -672,37 +672,6 @@ def msmarco_answerability_detection_to_tsv(dataDir, readFile, wrtDir, transParam
     devDf.to_csv(os.path.join(wrtDir, 'msmarco_answerability_test.tsv'), sep='\t', index=False, header=False)
     print('Test file written at: ', os.path.join(wrtDir, 'msmarco_answerability_test.tsv'))
     
-def query_correctness_to_tsv(dataDir, readFile, wrtDir, transParamDict, isTrainFile=False):
-
-    """
-
-    - Query correctness transformed file
-
-    For using this transform function, set ``transform_func`` : **query_correctness_to_tsv** in transform file.
-
-    Args:
-        dataDir (:obj:`str`) : Path to the directory where the raw data files to be read are present..
-        readFile (:obj:`str`) : This is the file which is currently being read and transformed by the function.
-        wrtDir (:obj:`str`) : Path to the directory where to save the transformed tsv files.
-        transParamDict (:obj:`dict`, defaults to :obj:`None`): Dictionary of function specific parameters. Not required for this transformation function.
-
-    """
-    print('Making data from file {}'.format(readFile))
-    df = pd.read_csv(os.path.join(dataDir, readFile), sep='\t', header=None, names = ['query', 'label'])
-    
-    # we consider anything above 0.6 as structured query (3 or more annotations as structured), and others as non-structured
-    
-    #df['label'] = [str(lab) for lab in df['label']]
-    df['label'] = [int(lab>=0.6)for lab in df['label']]
-    
-    data = [ [str(i), str(row['label']), row['query'] ] for i, row in df.iterrows()]
-    
-    wrtDf = pd.DataFrame(data, columns = ['uid', 'label', 'query'])
-    
-    #writing
-    wrtDf.to_csv(os.path.join(wrtDir, 'query_correctness_{}'.format(readFile)), sep="\t", index=False, header=False)
-    print('File saved at: ', os.path.join(wrtDir, 'query_correctness_{}'.format(readFile)))
-    
 def clinc_out_of_scope_to_tsv(dataDir, readFile, wrtDir, transParamDict, isTrainFile=False):
     
     """

Original file line number	Diff line number	Diff line change
`@@ -139,10 +139,10 @@`
`139`	`139`	`" --eval_batch_size 64 \\\n",`
`140`	`140`	`" --grad_accumulation_steps 1 \\\n",`
`141`	`141`	`" --log_per_updates 100 \\\n",`
`142`		`- " --eval_while_train True \\\n",`
`143`		`- " --test_while_train True \\\n",`
`144`	`142`	`" --max_seq_len 128 \\\n",`
`145`		`- " --silent True "`
	`143`	`+ " --eval_while_train \\\n",`
	`144`	`+ " --test_while_train \\\n",`
	`145`	`+ " --silent"`
`146`	`146`	`]`
`147`	`147`	`},`
`148`	`148`	`{`
Original file line number	Diff line number	Diff line change
`@@ -58,10 +58,10 @@`
`58`	`58`	`" --eval_batch_size 32 \\\n",`
`59`	`59`	`" --grad_accumulation_steps 1 \\\n",`
`60`	`60`	`" --log_per_updates 50 \\\n",`
`61`		`- " --eval_while_train True \\\n",`
`62`		`- " --test_while_train True \\\n",`
`63`	`61`	`" --max_seq_len 50 \\\n",`
`64`		`- " --silent True "`
	`62`	`+ " --eval_while_train \\\n",`
	`63`	`+ " --test_while_train \\\n",`
	`64`	`+ " --silent"`
`65`	`65`	`]`
`66`	`66`	`},`
`67`	`67`	`{`
Original file line number	Diff line number	Diff line change
`@@ -59,10 +59,10 @@`
`59`	`59`	`" --eval_batch_size 64 \\\n",`
`60`	`60`	`" --grad_accumulation_steps 1 \\\n",`
`61`	`61`	`" --log_per_updates 100 \\\n",`
`62`		`- " --eval_while_train True \\\n",`
`63`		`- " --test_while_train True \\\n",`
`64`	`62`	`" --max_seq_len 60 \\\n",`
`65`		`- " --silent True"`
	`63`	`+ " --eval_while_train \\\n",`
	`64`	`+ " --test_while_train \\\n",`
	`65`	`+ " --silent"`
`66`	`66`	`]`
`67`	`67`	`},`
`68`	`68`	`{`