Update llama_flash_attn_monkey_patch.py for flash attention 2

[yilin-bao]

Upgrading from FlashAttention (1.x) to FlashAttention-2

These functions have been renamed:

flash_attn_unpadded_qkvpacked_func -> flash_attn_varlen_qkvpacked_func

You can check how they made the update: https://github.com/Dao-AILab/flash-attention

Why are these changes needed?

The function flash_attn_unpadded_qkvpacked_func is not included anymore in
flash-attention, so causing issue with fine-tuning. This will fix the problem.

Related issue number (if applicable)

Checks

• I've run format.sh to lint the changes in this PR.
• I've included any doc changes needed.
  • https://github.com/Dao-AILab/flash-attention
• I've made sure the relevant tests are passing (if applicable).

[yilin-bao]

Also, could point out that fastchat is depending on flash-attention edition v1.0.9 or lower. That could also fix the problem.