Modelpublic

ghzlmc/gemma4-26b-a4b-unified

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes6downloads

Model Card

gemma4-26b-a4b-unified

Merged full-checkpoint export of the Gemma 4 26B A4B unified fine-tune.

Loading

python

from transformers import AutoModelForCausalLM, AutoProcessor

model_id = "ghzlmc/gemma4-26b-a4b-unified"
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True,
)

Training

Fine-tuned from unsloth/gemma-4-26b-a4b-it with TRL SFT and exported as merged safetensors shards.