mradermacher

/

model_requests

like 49

This is an uncensured version of Hermes-3-Llama-3.1-70B. It was finetuned on the Guilherme34/uncensor dataset using axolotl.

This is the first 70B model ever finetuned on my StormPeak node using 2x RTX 4090 GPUs. This was so thigh to fit into GPU memory but we somehow managed to make it fit allowing us to finetune 70B for free. No worries when I do finetunes I make sure to run them during nighttime and set the /tmp/pause flag to not affect your imatrix calculation tasks.