Search results

scheduler.py ZeroDivisionError when num_speculative_tokens=0 and method=mtp

Status: Open.

#604 In ROCm/ATOM;

· josusanmartin opened on Apr 19, 2026

[Perf Regression] 19 config(s) regressed @ 3138d9a4

performance

regression

Status: Open.

#603 In ROCm/ATOM;

· github-actions opened on Apr 19, 2026

[Issue]: RuntimeError with ATOM docker of rocm/atom:rocm7.2.1-ubuntu24.04-pytorch2.9.1-atom0.1.2 for ISL/OSL of 1k/1k and conc64 due to dtype mismatch of two matrices in one gemm kernel call.

ATOMAdded to frameworks-internal ATOM GitHub boardAdded to frameworks-internal ATOM GitHub board

Status: Open.

#577 In ROCm/ATOM;

· fululi12 opened on Apr 15, 2026

CI: Qwen3-Next-80B-A3B-Thinking accuracy test times out at 30-minute limit

ATOMAdded to frameworks-internal ATOM GitHub boardAdded to frameworks-internal ATOM GitHub board

Status: Open.

#565 In ROCm/ATOM;

· sunway513 opened on Apr 14, 2026

[parity with vllm & sglang]: forcing a specific acceptance rate during rejection sampling

Status: Open.

#555 In ROCm/ATOM;

· functionstackx opened on Apr 14, 2026

[Feature]: ROCm Kimi K2.5 EAGLE3 MTP heads

Status: Open.

#553 In ROCm/ATOM;

· functionstackx opened on Apr 14, 2026

[Feature]: docs on how to use atom with claude code

Status: Open.

#552 In ROCm/ATOM;

· functionstackx opened on Apr 14, 2026

RFC: Integrate rocm-trace-lite (RTL) for GPU kernel profiling in ATOM

Status: Open.

#534 In ROCm/ATOM;

· sunway513 opened on Apr 9, 2026

[Feature]: Make it so vLLM + OOT ATOM Plugin works with MiniMax 2.5

Status: Open.

#521 In ROCm/ATOM;

· rasmith opened on Apr 8, 2026

[Issue]: Multi-Stream Overlap Failure Due to HW Queue Collision

Status: Open.

#506 In ROCm/ATOM;

· ZhangLirong-amd opened on Apr 7, 2026

MXFP4 MoE weight quantization fails for MiniMax-M2.5: intermediate_size not divisible by block_n=128

Status: Open.

#490 In ROCm/ATOM;

· sunway513 opened on Apr 5, 2026

[Feature]: ATOM OpenAI Server, OpenClaw

Status: Open.

#488 In ROCm/ATOM;

· carlushuang opened on Apr 5, 2026