3 ms·
Typical LLM pretraining is very inefficient with data. NanoGPT slowrun shows that data can be used much more efficiently.
by naasking 2mo ago
Typical LLM pretraining is very inefficient with data. NanoGPT slowrun shows that data can be used much more efficiently.