diff --git a/WizardCoder/src/download_starcoder.py b/WizardCoder/src/download_starcoder.py
new file mode 100644
index 0000000..b09c052
--- /dev/null
+++ b/WizardCoder/src/download_starcoder.py
@@ -0,0 +1,15 @@
+model="bigcode/starcoder"
+
+cache_dir="/workspace/asr/WizardLM/WizardCoder"
+
+import transformers
+
+model = transformers.AutoModelForCausalLM.from_pretrained(
+    model,
+    cache_dir=cache_dir,
+    use_auth_token=True
+)
+
+print(model)
+
+print(sum(p.numel() for p in model.parameters()))
diff --git a/WizardCoder/src/train_wizardcoder.py b/WizardCoder/src/train_wizardcoder.py
index 04bbdc2..fd8bc0d 100644
--- a/WizardCoder/src/train_wizardcoder.py
+++ b/WizardCoder/src/train_wizardcoder.py
@@ -236,7 +236,7 @@ def train():
         for index in random.sample(range(len(train_dataset)), 3):
             print(f"Sample {index} of the training set: {train_dataset[index]}.")
     
-    data_collator = DataCollatorForSupervisedDataset(tokenizer=tokenizer)
+    data_collator = DataCollatorForSupervisedDataset(tokenizer=tokenizer) # DataCollatorForSupervisedDataset(tokenizer=GPT2TokenizerFast(name_or_path='bigcode/starcoder', vocab_size=49152, model_max_length=2048, is_fast=True, padding_side='right', truncation_side='right', special_tokens={'bos_token': '<|endoftext|>', 'eos_token': '<|endoftext|>', 'unk_token': '<|endoftext|>', 'pad_token': '[PAD]', 'additional_special_tokens': ['<|endoftext|>', '<fim_prefix>', '<fim_middle>', '<fim_suffix>', '<fim_pad>', '<filename>', '<gh_stars>', '<issue_start>', '<issue_comment>', '<issue_closed>', '<jupyter_start>', '<jupyter_text>', '<jupyter_code>', '<jupyter_output>', '<empty_output>', '<commit_before>', '<commit_msg>', '<commit_after>', '<reponame>']}, clean_up_tokenization_spaces=True))
     data_module = dict(train_dataset=train_dataset, eval_dataset=None, data_collator=data_collator)
 
     #Tell Trainer not to attempt DataParallel