BetEvening

• DeepSeek-V2-Lite Finetuning w/4x 3090s?

  Posted by BetEvening@reddit | LocalLLaMA | View on Reddit | 4 comments

• Just letting models talk to themselves for a bit makes them so much more accurate in math. No other model in the LMSYS Arena ever gets close.

  Posted by BetEvening@reddit | LocalLLaMA | View on Reddit | 12 comments