Merging duplicate functions in IncMHA, SpecIncMHA, and TreeIncNHA

[jiazhihao]

Description of changes:

This PR is an attempt to merge the duplicate functions and code across IncMHA, SpecIncMHA, and TreeIncMHA implementations.

Related Issues:

Linked Issues:

• Issue #

Issues closed by this PR:

• Closes #

Before merging:

• Did you update the flexflow-third-party repo, if modifying any of the Cmake files, the build configs, or the submodules?

[goliaro]

Looks good! This will help us remove the duplicated code. Feel free to merge!

[xinhaoc]

on my side the bug is fixed.

[jiazhihao]

@xinhaoc What's the command line you used to run SpecInfer? I am still observing random output tokens using the following command line:

./inference/spec_infer/spec_infer -ll:gpu 4 -ll:fsize 14000 -ll:zsize 30000 -llm-model llama -llm-weight ../inference/weights/llama_190M_weights/ -llm-config ../inference/models/configs/llama_190M.json -ssm-model llama -ssm-weight ../inference/weights/llama_190M_weights/ -ssm-config ../inference/models/configs/llama_190M.json -tokenizer ../inference/tokenizer/tokenizer.model -prompt ../inference/prompt/test.json