ekryski/diffusiongemma-26B-A4B-it-4bit
021
v2: fused experts (4-bit) + 4-bit embeddings, even ~3.84GB shards (FFAI ffai convert)
4-bit (affine, group_size 64) FFAI quant of google/diffusiongemma-26B-A4B-it
initial commit
v2: fused experts (4-bit) + 4-bit embeddings, even ~3.84GB shards (FFAI ffai convert)
4-bit (affine, group_size 64) FFAI quant of google/diffusiongemma-26B-A4B-it
initial commit