mirror of https://github.com/THUDM/ChatGLM-6B
Update main.py
update padding to left for padding_side='left' in Tokenizer setting. just making code to be consistent with tokenizer setting.pull/328/head
parent
4371f7a572
commit
5507f607be
|
@ -185,8 +185,8 @@ def main():
|
|||
labels = [-100] * context_length + input_ids[mask_position+1:]
|
||||
|
||||
pad_len = max_seq_length - len(input_ids)
|
||||
input_ids = input_ids + [tokenizer.pad_token_id] * pad_len
|
||||
labels = labels + [tokenizer.pad_token_id] * pad_len
|
||||
input_ids = [tokenizer.pad_token_id] * pad_len + input_ids
|
||||
labels = [tokenizer.pad_token_id] * pad_len + labels
|
||||
|
||||
model_inputs["input_ids"].append(input_ids)
|
||||
model_inputs["labels"].append(labels)
|
||||
|
|
Loading…
Reference in New Issue