Search results

[Bug] Prefill worker crashes with DeepEP 'invalid resource handle' on GB200 (job 5118)

Status: Open.

#195 In NVIDIA/srt-slurm;

· ishandhanani opened on Jun 5, 2026

Missing sa-bench deps - ensurepip not available

Status: Open.

#188 In NVIDIA/srt-slurm;

· faradawn opened on May 29, 2026

Cannot Sweep EP values for vLLM

Status: Open.

#186 In NVIDIA/srt-slurm;

· faradawn opened on May 29, 2026

[Bug] decode workers CUDA OOM due to mem-fraction-static 0.95 on GB200 GPUs

Status: Open.

#172 In NVIDIA/srt-slurm;

· ishandhanani opened on May 21, 2026

[Bug] Prefill worker crashes with DeepEP 'invalid resource handle' after DeepGEMM warmup

Status: Open.

#171 In NVIDIA/srt-slurm;

· ishandhanani opened on May 21, 2026

default_bash_preamble silently skipped when use_bash_wrapper=False (e.g. nginx)

Status: Open.

#107 In NVIDIA/srt-slurm;

· ishandhanani opened on Apr 28, 2026

Port GPQA / MMLU / longbenchv2 / GSM8K to the configs/<bench>/ script pattern (deprecate structured runners)

Status: Open.

#92 In NVIDIA/srt-slurm;

· ishandhanani opened on Apr 27, 2026

GB300 poor evals on InferenceX

Status: Open.

#51 In NVIDIA/srt-slurm;

· Oseltamivir opened on Apr 21, 2026

Decode workers crash with NVFP4 dispatch scale tensor assertion

Status: Open.

#46 In NVIDIA/srt-slurm;

· ishandhanani opened on Apr 18, 2026

[Bug] DeepEP assertion failure due to insufficient SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK

Status: Open.

#44 In NVIDIA/srt-slurm;

· ishandhanani opened on Apr 17, 2026

DeepEP assertion failure during decode worker warmup: num_max_dispatch_tokens_per_rank too small

Status: Open.

#42 In NVIDIA/srt-slurm;

· ishandhanani opened on Apr 17, 2026