Modelpublic

uwuwuwuwuwuwu/kimi-k2.5-reward-hacking-step-560

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes9downloads

Model Card

Tinker LoRA Adapter

This repository contains a LoRA adapter exported from Tinker.

Usage

python

from transformers import AutoModelForCausalLM

adapter_id = "uwuwuwuwuwuwu/kimi-k2.5-reward-hacking-step-560"
base_model = "moonshotai/Kimi-K2.5"

model = AutoModelForCausalLM.from_pretrained(adapter_id, device_map="auto")

Source

tinker://bdb75fef-52ff-5675-891b-f915753a3117:train:0/sampler_weights/000096

Details

—Base model: moonshotai/Kimi-K2.5
—LoRA rank: 32
—Trained modules: attn=True, mlp=True, unembed=True