v0.1.9a4

What's Changed

• [model] fix: qwen3_moe router double-softmax and idempotent _init_weights wrap by @TimYangst in #715
• [ops, model] feat: dispatch Qwen3.5 linear-attention kernels via OpSlot by @TimYangst in #714
• [ops, model] feat: forward temperature and return entropy from log-probs path by @Luosuu in #720
• [ci, model] test: add bitwise logits-equal tests for transformers v4 models by @TimYangst in #721
• [release] chore: release v0.1.9a4 by @Luosuu in #723

Full Changelog: v0.1.9a3...v0.1.9a4