Qlora + axolotl + good foundation model (llama/mistral/etc, usually instruction fine tuned) + runpod works great.
A single A100 or H100 with 80GB VRAM can fine tune 70B open models (and obviously scaling out to many nodes/GPUs is faster, or can use much cheaper GPUs for fine tuning smaller models.)
A single A100 or H100 with 80GB VRAM can fine tune 70B open models (and obviously scaling out to many nodes/GPUs is faster, or can use much cheaper GPUs for fine tuning smaller models.)
The localllama Reddit sub at https://www.reddit.com/r/LocalLLaMA/ is also an awesome community for the GPU poor :)