commit from tesla

app.py

@@ -1,5 +1,5 @@
 import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, TrainingArguments, Trainer, BertLMHeadModel, BertForSequenceClassification
+from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig, TrainingArguments, Trainer, BertLMHeadModel, BertForSequenceClassification
 from datasets import Dataset
 import pandas as pd
 import csv
@@ -67,6 +67,11 @@ def main(api_name, base_url):
     device = get_device()  # Get the appropriate device
     data = load_data_and_config("train2.csv")
     tokenizer = AutoTokenizer.from_pretrained("google/codegemma-2b")
+    # Load the configuration for a specific model
+    config = AutoConfig.from_pretrained('google/codegemma-2b')
+    # Update the activation function
+    config.hidden_act = 'gelu_pytorch_tanh'  # Set to use approximate GeLU
+    
     model = AutoModelForCausalLM.from_pretrained('google/codegemma-2b', is_decoder=True)
     #model = BertLMHeadModel.from_pretrained('google/codegemma-2b', is_decoder=True)
     # Example assuming you have a prepared dataset for classification
@@ -74,7 +79,8 @@ def main(api_name, base_url):
     model.to(device)  # Move model to the appropriate device
 
     train_model(model, tokenizer, data, device)
-
+ 
+    
     model.save_pretrained("./fine_tuned_model")
     tokenizer.save_pretrained("./fine_tuned_model")