[Feature] Optimize host-device sync problem in prefill phase for Qwen3Next/Qwen3.5 (#7967)

[Ascend950][quant][Feature] Add W4A4 MXFP4 quantization support for Ascend950 (#7877)

[BugFix][xlite] Clamp block_size to 128 for xlite graph compatibility (#7943)

[CI][Lint] Restrict python-init check to tracked package directories (#7939)

[CI] Main2main upgrade to 0324 (#7787)

[BugFix] fix qwen3-next compilation error (#7936)

[Feature] Add end-to-end release management skill (#6971)

[CI] add nightly Qwen3.5_27B_w8a8;MiniMax-M2.5-w8a8;Qwen3.5-397B-w8a8-mtp (#7745)

[CI] fix acc baseline of qwen3vl 235b (#7982)

[CI] add qwen3vl weights download (#7917)

[CI] add nightly ci test for qwen3vl (#7914)

[Feature] Support Flash Comm V1 for Qwen3-VL models (#7897)

[Feature] [310p] Add recurrent_gated_delta_rule_310 AscendC Custom Op (#7926)

[Doc][Misc] Add GLM5 to supported model list and update deployment document for GLM5 (#7958)

[Performance][model_runner_v2]:optimize the performance of the _ranks_kernel and _min_p_kernel (#7767)