Update README.md

README.md

@@ -88,12 +88,67 @@ Use the code below to get started with the model.
 #### Preprocessing [optional]
 
 [More Information Needed]
+train_df['emotion_stance'] = "Classify based on the features:" + train_df['target_emotion_stance'].apply(lambda x: str(x)) + " in the text: "  + train_df['source_text']
+train_df['emotion'] = "Classify based on the features:" + train_df['target_emotion'].apply(lambda x: str(x)) + " in the text: "  + train_df['source_text']
+train_df['stance'] = "Classify based on the features:" + train_df['target_stance'].apply(lambda x: str(x)) + " in the text: "  + train_df['source_text']
+
+test_df['emotion_stance'] = "Classify based on the features:" + test_df['target_emotion_stance'].apply(lambda x: str(x)) + " in the text: "  + test_df['source_text']
+test_df['emotion'] = "Classify based on the features:" + test_df['target_emotion'].apply(lambda x: str(x)) + " in the text: "  + test_df['source_text']
+test_df['stance'] = "Classify based on the features:" + test_df['target_stance'].apply(lambda x: str(x)) + " in the text: "  + test_df['source_text']
+
+val_df['emotion_stance'] = "Classify based on the features:" + val_df['target_emotion_stance'].apply(lambda x: str(x)) + " in the text: "  + val_df['source_text']
+val_df['emotion'] = "Classify based on the features:" + val_df['target_emotion'].apply(lambda x: str(x)) + " in the text: "  + val_df['source_text']
+val_df['stance'] = "Classify based on the features:" + val_df['target_stance'].apply(lambda x: str(x)) + " in the text: "  + val_df['source_text']
+
+#label_dict = {0:'FAKE', 1:'REAL'}
+
+train_df['target_text'] = train_df['target_text'].apply(lambda x: str(x))
+val_df['target_text'] = val_df['target_text'].apply(lambda x: str(x))
+test_df['target_text'] = test_df['target_text'].apply(lambda x: str(x))
+
+max_length = 500
+
+train_df = train_df[['source_text', 'target_text', #'src', #'dict_target_emotion_stance',
+                      'target_emotion_stance', 'target_emotion', 'target_stance',
+                      'emotion_stance', 'emotion', 'stance']]
+
+test_df = test_df[['source_text', 'target_text', #'src', #'dict_target_emotion_stance',
+                      'target_emotion_stance', 'target_emotion', 'target_stance',
+                      'emotion_stance', 'emotion', 'stance']]
+
+val_df = val_df[['source_text', 'target_text', #'src', #'dict_target_emotion_stance',
+                      'target_emotion_stance', 'target_emotion', 'target_stance',
+                      'emotion_stance', 'emotion', 'stance']]
 
 
 #### Training Hyperparameters
 
 - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
 
+TrainingArguments(
+    output_dir="temp",
+    evaluation_strategy="epoch",
+    learning_rate=1e-3,
+    gradient_accumulation_steps=1,
+    #auto_find_batch_size=True,
+    num_train_epochs=2,
+    #save_steps=100,
+    weight_decay=0.01,
+    save_total_limit=2,
+
+
+    #optim="adafactor",
+    optim="adamw_torch_fused",
+    per_device_train_batch_size=128,
+    per_device_eval_batch_size=128,
+
+    # save_steps=1000,
+    # evaluation_strategy ="steps",
+    metric_for_best_model = 'eval_loss',  #eval_loss
+    save_strategy="epoch",
+    load_best_model_at_end=True,
+)
+
 #### Speeds, Sizes, Times [optional]
 
 <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->