Audio

Lex Fridman Podcast2024-09

Latent Space Podcast2023-07

Longer

Llama-2 is was expensive2023-08

Shorter

4-bit weight-quantization is more expensive than 16-bit2023-08

flash attention isn't helpful when generating tokens2023-05

llama-1 needs multi-query attention2023-05

instruction finetuning is underrated2023-04