Try with the original gptj checkpoint

app.py

@@ -3,19 +3,20 @@ import torch
 from transformers import pipeline, GPTJForCausalLM, AutoModelForCausalLM
 from peft import LoraConfig, get_peft_model, PeftModel, PeftConfig
 
-config = PeftConfig.from_pretrained("hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es")
-model = AutoModelForCausalLM.from_pretrained("hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es", return_dict=True, load_in_8bit=True, device_map='auto')
+# config = PeftConfig.from_pretrained("hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es")
+# model = AutoModelForCausalLM.from_pretrained("hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es", return_dict=True, load_in_8bit=True, device_map='auto')
 
-# # load fp 16 model
-# model = GPTJForCausalLM.from_pretrained("models/hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es")
+# # load tokenizer
+# tokenizer = AutoTokenizer.from_pretrained("hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es")
+
+# # Load the Lora model
+# model = PeftModel.from_pretrained(model, "hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es")
 
-# config = AutoConfig.from_pretrained("hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es", name_or_path="adapter_model.bin")
 
-# load tokenizer
-tokenizer = AutoTokenizer.from_pretrained("hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es")
+# # load fp 16 model
+model = GPTJForCausalLM.from_pretrained("bertin-project/bertin-gpt-j-6B")
 
-# Load the Lora model
-model = PeftModel.from_pretrained(model, "hackathon-somos-nlp-2023/bertin-gpt-j-6b-ner-es")
+config = AutoConfig.from_pretrained("bertin-project/bertin-gpt-j-6B")
 
 # create pipeline
 pipe = pipeline("text-generation", model=model, config=config, tokenizer=tokenizer, device=0,)