remove llama 70b (#21396)
Remove llama 70b model due to security reason.
We need add shard code in HF to enable model shardding for llama-70b,
these codes are not merged into main branch as HF forks want a more
general solution instead of doing shard for specify model. shared code
is kept here:
https://github.com/frank-dong-ms/transformers/tree/frdong/shard_llama
we kept llama-70b related code here for internal use:
https://github.com/frank-dong-ms/onnxruntime/tree/frdong/llama_70b
