gemma4-zero-compute

Optimized Gemma4 text-only checkpoint exported from this repository.

Loading

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("haysonC/gemma4-zero-compute", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("haysonC/gemma4-zero-compute", trust_remote_code=True)

Notes

• Base model: google/gemma-4-26B-A4B-it
• Architecture: OptimizedGemma4ForCausalLM
• Scope: text-only causal language model export
• Router checkpoint loaded: True