[Frontend] Allo-PyTorch Enhancement

[chhzh123]

Currently, Allo has already had a compilation flow from PyTorch, but the generated program is not yet high-performance. We need to enhance automatic schedule composition and incorporate linalg/tensor dialects for better representing the neural network operations.

[LarryFan]

10/21 Update

• Busy past 2 weeks with projects, exams, etc.
• Installed, Torch, Transformers to run example models
• Ran bert, toy, self-attention, mlp
• Recorded + Analyzing graphs and ASTs

[LarryFan]

11/4

Debugged MLIR and Vitis HLS to run benchmarks
Working on loop unroll and pipelining integration + test

[chhzh123]

@LarryFan Please check #248 that I've exposed the HLS target to the frontend