lingjiew93

How to solve the error(4103) when profling LLM training with MI250?

1

I'm running LLM training with MI250. The instruction and code I used are https://www.mosaicml.com/blog/amd-mi250 and https://github.com/mosaicml/llm-foundry It runs well without profiling, but when I tried to profile below errors are...

TCC_HIT_sum for MI210 is only half of MI100 when L2CacheHit is ~100%

3

Hi, I found that TCC_READ_sum is only half about the real size which from MI100 result. The same as TCC_HIT_sum. They both have 64 bytes cacheline size so the number...