Quantize Gemma4 per-layer embedding tables for WebGPU (INT4/INT8 configurable) - #403

Draft

feich-ms wants to merge 6 commits into

mainfrom

user/feich/quantize_per_layer_embedding_table

Commits on Jul 14, 2026

Quantize per-layer embedding tables to INT4 for WebGPU

feich-ms
and
claude
committed
Make embedding quantization bits configurable via embedding_bits param

feich-ms
and
claude
committed
Address PR review: validation, error handling, docstring, and tests

feich-ms
and
claude
committed

Commits on Jul 29, 2026