Search results

[bug] test_attention_decode_bf16 can't pass with block_size=32

Status: Open.

#26 In Tencent/hpc-ops;

· WingEdge777 opened on Feb 6, 2026

FP4 / NVFP4 / MXFP4 support

Status: Open.

#25 In Tencent/hpc-ops;

· arcticgus opened on Feb 5, 2026

vllm integrations

Status: Open.

#24 In Tencent/hpc-ops;

· qibaoyuan opened on Feb 5, 2026

test_fused_moe_blockwise.py 改了一下shape，单测过不了

Status: Open.

#19 In Tencent/hpc-ops;

· K11OntheBoat opened on Feb 2, 2026

Sglang integrations

Status: Open.

#16 In Tencent/hpc-ops;

· Kamizm opened on Jan 30, 2026

[Feature Request] Support for additional head_dim configurations (e.g., 64, 96)

Status: Open.

#11 In Tencent/hpc-ops;

· JiaxiangZheng opened on Jan 27, 2026

[Feature Request] Support q_len > 1 for FP8 decode (Multi-Token Prediction)

Status: Open.

#4 In Tencent/hpc-ops;

· ZZBoom opened on Jan 24, 2026