Create README.md

README.md

@@ -0,0 +1,15 @@
+training_args = TrainingArguments(
+    output_dir='./results',          # output directory
+    num_train_epochs=1,              # total number of training epochs
+    per_device_train_batch_size=8,  # batch size per device during training
+    #per_device_eval_batch_size=2,   # batch size for evaluation
+    warmup_steps=3000,                # number of warmup steps for learning rate scheduler (used to be 500)
+    weight_decay=0.01,               # strength of weight decay
+    learning_rate=0.1e-3, # default = 5e-5=0.5e-4
+    logging_dir='./logs',            # directory for storing logs
+    logging_steps=50,
+    #eval_steps = 100,
+    overwrite_output_dir = True,
+    save_strategy = 'epoch',
+    #logging_strategy = 'epoch',
+)