systalyze/gemma4-26b-a4b-it-dspark

Load via AutoModelForCausalLM; keep text-generation pipeline tag

Map AutoModelForCausalLM so Hub shows Text Generation

Add serving evals vs base and improve discoverability

Update README.md

Update README.md

Remove Evaluation section

Document required attn_implementation="flex_attention"

Rewrite model card for public release

Release step_7139 as v1.0

Retitle card, correct training status, add LICENSE and NOTICE

Add DSpark gemma-4-26B-A4B drafter, checkpoint step_7139 (step 7,139/25,960, epoch 3/10)

initial commit