Fix max_length issue

More info: huggingface/transformers#1791
0xsuid · Sep 24, 2022 · fcb961e · fcb961e
1 parent 7845eb7
commit fcb961e
Showing 1 changed file with 2 additions and 0 deletions.
diff --git a/gpt_neo_125m/tune_gpt.py b/gpt_neo_125m/tune_gpt.py
@@ -51,6 +51,8 @@ def format_input(dataset):
 coding_problems = format_input(raw_ds)
 
 max_length = max([len(tokenizer.encode(coding_problem)) for coding_problem in coding_problems])
+model_max_length = model.config.max_position_embeddings
+max_length = max_length if max_length <= model_max_length else model_max_length
 print("Max length: {}".format(max_length))