Ascend qwen3.6 optimizate baseline - #301

Open

appleinsky wants to merge 22 commits into

flagos-ai:v0.1.0from

appleinsky:add-qwen3_6_pto_chunk_gdn

Commits on Jun 24, 2026

feat(ascend): add split_qkv_rmsnorm_mrope fused kernel for Qwen3.5/3.6
appleinsky
committed

Commits on Jun 25, 2026

Integrate Ascend NPU graph backend and npugraph_ex passes into vllm-plugin-FL
appleinsky
committed
Make FL OOT ops Dynamo-safe on Ascend NPU for graph-mode serving
appleinsky
committed

Commits on Jul 1, 2026

feat(ascend): enable MoE MTP for Qwen3.6-35B-A3B
appleinsky
committed
fix(ascend): fallback to npu_fusion_attention for unsupported head_dim (e.g. 256)
appleinsky
committed

Commits on Jul 20, 2026

Commits on Jul 22, 2026

Commits on Jul 23, 2026

Merge pull request #4 from Lntano061105/optimize-qwen3_6-full-graph-decode
appleinsky
authored