Update finetune.py

finetune.py

@@ -17,7 +17,7 @@ from peft import (
 
 
 # Parameters
-MICRO_BATCH_SIZE = int(32)
+MICRO_BATCH_SIZE = 16
 BATCH_SIZE = 16
 size = "7b"
 GRADIENT_ACCUMULATION_STEPS = BATCH_SIZE // MICRO_BATCH_SIZE
@@ -152,6 +152,8 @@ trainer = transformers.Trainer(
         save_total_limit=100,
         load_best_model_at_end=True if VAL_SET_SIZE > 0 else False,
         ddp_find_unused_parameters=False if ddp else None,
+        push_to_hub=True,
+        push_to_hub_model_id="gustavoaq/llama_ft"
     ),
     data_collator=transformers.DataCollatorForLanguageModeling(tokenizer, mlm=False),
 )
@@ -163,5 +165,6 @@ model.state_dict = (
 ).__get__(model, type(model))
 
 trainer.train()
+trainer.push_to_hub()
 
 model.save_pretrained(OUTPUT_DIR)