B300 single node / B300 单节点

Open

#135 opened on Oct 22, 2025 4 comments 0 reactions 1 assignee View on GitHub

@kedarpotdar-nv is already working on this.

Since Oct 22, 2025.

Assessment

This issue has not been assessed yet.

Description

NVIDIA p1

In order to provide a baseline for B300 performance, add single node for GPT OSS and DeepSeek R1.

This will require:

Adding single node path to runners/launch_b300-nv.sh
Adding configs to nvidia-master.yaml
Adding appropriate benchmarks/ scenarios (FP4, FP8, MTP, etc) for GPT OSS and DeepSeek, vLLM, SGLang and TRT (for both)

中文说明

MI300X vLLM DeepSeek-R1 有 FP8 的 MTP 基准测试。

Dominant language: Python
Stars: 1.7k
Forks: 303
Avg merge: 1d 13h
Merged PRs (30d): 284

Contributor guide

Open the contributing guide

First steps

Read the whole issue, then the project's contributing guide.
Comment on the issue to say you are picking it up — it saves two people doing the same work.
Fork the repository and make your change on a branch.
Open a pull request that references the issue number.

More from SemiAnalysisAI/InferenceX

Removal of "1 The ATOM engine is promising, however it has yet to serve production tokens. It is still in its infant stage." Open

Difficulty 2/5 1-3 hours Newbie friendliness 68/100

SemiAnalysisAI/InferenceX#2125 · Opened Jul 9, 2026
[vLLM] Missing --max-model-len guard for offload-disabled agentic runs causes KV cache OOM on H100 / [vLLM] 禁用 offload 的 agentic 运行缺少 --max-model-len 保护导致 H100 上 KV 缓存 OOM Open

Difficulty 2/5 1-3 hours Newbie friendliness 68/100

SemiAnalysisAI/InferenceX#1587 · Opened May 29, 2026
[experimental chore/agentx-v0.2-aiperf-testing branch] agentic build_replay_cmd: missing --tokenizer-trust-remote-code breaks Kimi-K2.5 (and other custom-code tokenizer models) / agentic build_replay_cmd 缺少 --tokenizer-trust-remote-code 导致 Kimi-K2.5 等自定义 tokenizer 模型失败 Open

Difficulty 1/5 Under an hour Newbie friendliness 78/100

SemiAnalysisAI/InferenceX#1369 · 3 comments · Opened May 13, 2026
[experimental chore/agentx-v0.2-aiperf-testing branch] agentic build_replay_cmd: --cache-bust system_prefix mismatched with current aiperf scenario (requires first_turn_prefix) / agentic build_replay_cmd 的 --cache-bust system_prefix 与当前 aiperf 场景不匹配（需要 first_turn_prefix） Open

Difficulty 1/5 1-3 hours Newbie friendliness 76/100

SemiAnalysisAI/InferenceX#1359 · 1 comment · Opened May 13, 2026
[Question] Is hardware contribution to the official CI cluster supported? Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

SemiAnalysisAI/InferenceX#3122 · 3 comments · Opened Sep 15, 2026

All issues in SemiAnalysisAI/InferenceX

Similar issues

[Bug]: notebooks/06_visualization_quickstart.ipynb fails: SHAPVisualizer.plot_waterfall raises IndexError Open

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

bancolombia/sentinel#23 · Opened Sep 20, 2026
test md Open

CI

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

sympy/sympy#30547 · 1 comment · Opened Sep 20, 2026
quickjs: support `timeout=None` (no per-call deadline) in `CodeInterpreterMiddleware` Open

integration:quickjs org:external priority:backlog topic:code-interpreter topic:middleware type:feature

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

langchain-ai/deepagents#6450 · Opened Sep 20, 2026
OpenAISamplingHandler drops ToolResultContent when a user message's list content also carries text/image blocks Open

bug client

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

PrefectHQ/fastmcp#5182 · 1 comment · Opened Sep 20, 2026
last-mile drop-off on this SKILL.md Open

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

wshobson/agents#717 · Opened Sep 20, 2026

More Python issues

A short digest of beginner-friendly GitHub issues.