Modelpublic

jackzhang/openlm_3b_201305_dolci_think_300k_pre2013_sft_full-15500

sourceHugging Faceupdated 5mo agoView on Hugging Face

0likes9downloads

Model Card

openlm3b201305dolcithink300kpre2013sftfull-15500

SFT checkpoint (step 15500) of `dogtooth/open-lm-3b-201305`, trained on the Dolci-Think 300K dataset filtered to pre-2013 content to preserve the 201305 knowledge cutoff.

—Base model: dogtooth/open-lm-3b-201305
—Architecture: OpenLM (custom OpenLMForCausalLM, loaded via trust_remote_code=True)
—Training: Full SFT via LLaMA-Factory + DeepSpeed
—Checkpoint step: 15500

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "jackzhang/openlm_3b_201305_dolci_think_300k_pre2013_sft_full-15500"
tok = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id, trust_remote_code=True, dtype="bfloat16", device_map="auto"
)