Update README.md

README.md

@@ -18,6 +18,30 @@ Opt-6.7B-lora-caramelo is further pre-train [Facebook's OPT-6.78](https://huggin
 This model is intended to be used with fine-tuning, supervision, and/or moderation. 
 I recommend having a human curate or filter the outputs.
 
+
+### How to use
+
+```py
+import torch
+from peft import PeftModel, PeftConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+peft_model_id = "arthurangelici/opt-6.7b-lora-caramelo"
+config = PeftConfig.from_pretrained(peft_model_id)
+model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=True, device_map='auto')
+tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
+
+# Load the Lora model
+model = PeftModel.from_pretrained(model, peft_model_id)
+
+batch = tokenizer("Caramelo é um simbolo do: ", return_tensors='pt')
+
+with torch.cuda.amp.autocast():
+  output_tokens = model.generate(**batch, max_new_tokens=50)
+
+print('\n\n', tokenizer.decode(output_tokens[0], skip_special_tokens=True))
+```
+
 ### License
 
 The model is licensed under the OPT-6.75B license, Copyright (c) Meta Platforms, Inc. All Rights Reserved.