Upload code/ with huggingface_hub

code/inference.py

@@ -278,7 +278,7 @@ JSON_CONTENT_TYPE = 'application/json'
 # loads the model into memory from disk and returns it
 def model_fn(model_dir):
   # Load model from HuggingFace Hub
-  tokenizer = T5TokenizerFast(model_dir, extra_ids=0)
+  tokenizer = T5TokenizerFast.from_pretrained(model_dir, extra_ids=0)
   model = AutoModelForSeq2SeqLM.from_pretrained(model_dir)
   return model, tokenizer