Load model in 4bit with bitsandbytes

Add requirements.txt with bitsandbytes

Use LlamaTokenizerFast in handler

Use LlamaTokenizerFast

Remove backend field from tokenizer config

Fresh handler with LlamaTokenizer

Delete old handler

Update tokenizer_config.json

Update tokenizer_config.json

Create handler.py

Upload folder using huggingface_hub

initial commit