Search results

how to pre-train qformer?

Status: Open.

#60 In xiaomi-mlab/Orion;

· Yzichen opened on Mar 25, 2026

bench2drive联调

Status: Open.

#57 In xiaomi-mlab/Orion;

· guangyuanQ opened on Jan 30, 2026

训练阶段划分问题

Status: Open.

#56 In xiaomi-mlab/Orion;

· phr-dlut opened on Jan 23, 2026

实际测试结果与论文中的结果差距很大

Status: Open.

#55 In xiaomi-mlab/Orion;

· yabuke opened on Jan 18, 2026

Question about initialization for the Qwen2-0.5B experiment in the appendix

Status: Open.

#54 In xiaomi-mlab/Orion;

· dream12251108 opened on Jan 15, 2026

no prepare_B2D.py

Status: Open.

#53 In xiaomi-mlab/Orion;

· leoyang-go opened on Jan 12, 2026

DIFF模式训练遇到的问题

Status: Open.

#52 In xiaomi-mlab/Orion;

· q7jin opened on Jan 6, 2026

A question about NC

Status: Open.

#51 In xiaomi-mlab/Orion;

· sonnefred opened on Dec 19, 2025

How much GPU memory used in stage 1 traning?

Status: Open.

#50 In xiaomi-mlab/Orion;

· cuge1995 opened on Dec 17, 2025

How to output the predicted trajectory

Status: Open.

#47 In xiaomi-mlab/Orion;

· sonnefred opened on Nov 27, 2025

Model and python process crashed

Status: Open.

#43 In xiaomi-mlab/Orion;

· wzh506 opened on Nov 6, 2025

Conflict between flash_atten and cuda112 in h20 training

Status: Open.

#41 In xiaomi-mlab/Orion;

· Ansel12306 opened on Oct 25, 2025