Search results

rocm/atom-dev:latest crashes on MiniMax-M3 MXFP4: AIter API mismatch in fused_qknorm_idxrqknorm

Status: Open.

#1347 In ROCm/ATOM;

· Princejain1101 opened on Jun 24, 2026

[Perf Regression] 16 config(s) regressed @ ef446036

performance

regression

Status: Open.

#1338 In ROCm/ATOM;

· github-actions opened on Jun 24, 2026

[Perf Regression] 20 config(s) regressed @ 345d6a53

performance

regression

Status: Open.

#1327 In ROCm/ATOM;

· github-actions opened on Jun 23, 2026

[BUG] Performance degradation and garbled text output when benchmarking DeepSeek-V4-Pro with ATOM backend on MI355X

Status: Open.

#1302 In ROCm/ATOM;

· luqitao opened on Jun 21, 2026

[Perf Regression] 36 config(s) regressed @ d7964d50

performance

regression

Status: Open.

#1296 In ROCm/ATOM;

· github-actions opened on Jun 20, 2026

[Perf Regression] 26 config(s) regressed @ e79fe6f5

performance

regression

Status: Open.

#1251 In ROCm/ATOM;

· github-actions opened on Jun 17, 2026

[Issue]: Output degeneration on second identical request with prefix caching (MiniMax M2.5, MI350X)

Status: Open.

#1239 In ROCm/ATOM;

· jamesETsmith opened on Jun 16, 2026

[Issue]: Running Minimax 2.5/2.7 using tip of tree (ATOM/AITER/vLLM) or latest nightly version with TP=8 on results in error: Invalid qk hidden dim layout Invalid qk hidden dim layout for qknorm_allreduce_fusion_kernel_2stage kernel:

Status: Open.

#1223 In ROCm/ATOM;

· rasmith opened on Jun 15, 2026

[Issue]: Using basic instructions to run simple example results in crash on gfx950

Status: Open.

#1207 In ROCm/ATOM;

· rasmith opened on Jun 13, 2026

Real World ATOM Kimi K2.7-Code Production Inferencing Enablement Request

Status: Open.

#1188 In ROCm/ATOM;

· functionstackx opened on Jun 12, 2026

[Parity with CUDA vLLM, CUDA SGLang]: native ATOM llm-d Kubernetes Distributed Inferencing Orchestrator for Production LLM Inferencing [Functional Enablement + Guides + Nightly CI]

Status: Open.

#1187 In ROCm/ATOM;

· functionstackx opened on Jun 12, 2026

[Parity with CUDA vLLM]: ATOM Nightly CI and/or per PR CI Function Calling Eval

Status: Open.

#1186 In ROCm/ATOM;

· functionstackx opened on Jun 12, 2026