sammysun0711

[sammysun0711]

Details:
Team: sammysun0711
Leaderboard score (total speedup): 15.0266

1. Add additional error message and performance evaluation latency in memory for each agent code generation iteration
2. Modify prompt_for_generation and prompt_for_reflection to add target platform MI300X GPU hardware information for triton kernel auto tuning
3. Try to dumpy memory state into disk for further fine tuning
4. Upload generated 10 Triton kernel and performance benchmark & accuracy check results on target MI300X GPU.
5. Minor: add pycache to .gitignore