jnjj/gemma-3-1b-it-qat-int4-quantized-inference-unrestricted-pruned-sf

Create README.md

Upload INT4 quantized Gemma‑3‑1B‑IT QAT with bfloat16 compute, 90% magnitude pruning, extensive unconventional modifications including instruction conversion and GPTQ/AutoGPTQ flags, and only bfloat16 .weight tensors saved as safetensors (bfloat16 compute)

initial commit