Modelpublic

xiaolin-july/olmo3-190m-zh-nano-continue

sourceHugging Faceapache-2.0updated 3mo agoView on Hugging Face

0likes9downloads

Model Card

xiaolin-july/olmo3-190m-zh-nano-continue

持续预训练版本：基于 xiaolin-july/olmo3-190m-zh-nano，在 cmz1024/llm101-olmo3-zh-demo-data/continue 数据上继续训练。

训练配置

—数据：cmz1024/llm101-olmo3-zh-demo-data/continue/full_v31.bin
—LR：2e-4（低 LR 防止灾难性遗忘）
—Warmup：10%

用法

python

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("xiaolin-july/olmo3-190m-zh-nano-continue")
tok = AutoTokenizer.from_pretrained("xiaolin-july/olmo3-190m-zh-nano-continue")