Add comprehensive model card with dataset tag, benchmarks, and usage examples

Add tokenizer

Add fine-tuned DeBERTa-v3-small

initial commit