Search results

Kernel recompiles when the token count changes

Status: Open.

#60 In Dao-AILab/sonic-moe;

· ad8e opened on May 23, 2026

Support for PyTorch 2.10

Status: Open.

#58 In Dao-AILab/sonic-moe;

· LivingFutureLab opened on May 12, 2026

TypeError: a_dtype should be float16 or float8

Status: Open.

#56 In Dao-AILab/sonic-moe;

· conceptofmind opened on May 10, 2026

Does sonic-moe plan to support mxfp4 w4a16?

Status: Open.

#53 In Dao-AILab/sonic-moe;

· yuan-luo opened on Apr 29, 2026

Sentinel experts support for EP

Status: Open.

#50 In Dao-AILab/sonic-moe;

· IlyasMoutawwakil opened on Apr 24, 2026

plan for FP8 support

Status: Open.

#28 In Dao-AILab/sonic-moe;

· tanyuxin-aaa opened on Feb 10, 2026

No import guidance in README

documentationImprovements or additions to documentationImprovements or additions to documentation

enhancementNew feature or requestNew feature or request

Status: Open.

#17 In Dao-AILab/sonic-moe;

· clocksmith opened on Jan 13, 2026

token_gather_sum_kernel cost too much time

enhancementNew feature or requestNew feature or request

Status: Open.

#14 In Dao-AILab/sonic-moe;

· attack204 opened on Jan 1, 2026

Megatron integration

enhancementNew feature or requestNew feature or request

Status: Open.

#7 In Dao-AILab/sonic-moe;

· Birch-san opened on Dec 23, 2025