Issues

• Training cost

  inclusionAI/TwinFlow#34 · Filter by author TensorsSun opened on Sep 21, 2026

• 关于数据配比的问题

  inclusionAI/TwinFlow#33 · Filter by author Zeth-Lee opened on Sep 9, 2026

• Why does the LoRA variant use t - tt/2 to approximate the two time embedders instead of adding an auxiliary time embedder?

  inclusionAI/TwinFlow#32 · Filter by author Eeelyyy opened on Aug 25, 2026

• Verify evals on Papers with Code

  inclusionAI/TwinFlow#31 · Filter by author NielsRogge opened on Aug 16, 2026

• Qwen-Image lora 2步/4步效果不佳

  inclusionAI/TwinFlow#30 · Filter by author pjyang-zju-1 opened on Jun 8, 2026

• 只用TwinFlow Loss是否能够约束模型进行任意步生成

  inclusionAI/TwinFlow#29 · Filter by author M45hiro opened on May 28, 2026
  1

• 关于时间尺度的问题

  inclusionAI/TwinFlow#28 · Filter by author M45hiro opened on Apr 24, 2026
  1

• Why 20B models need fewer steps than smaller ones?

  inclusionAI/TwinFlow#27 · Filter by author vitrun opened on Apr 7, 2026
  1

• SD3 inference code

  inclusionAI/TwinFlow#26 · Filter by author Joyies opened on Mar 26, 2026

• How about the effect of L_twin only?

  inclusionAI/TwinFlow#25 · Filter by author toyot-li opened on Feb 23, 2026
  1

• QR code is invalid

  inclusionAI/TwinFlow#24 · Filter by author toyot-li opened on Feb 23, 2026
  3

• qwen edit 2509 训练lora效果很差

  inclusionAI/TwinFlow#23 · Filter by author hello-program opened on Jan 22, 2026
  6

• batch size 能否为1？

  inclusionAI/TwinFlow#19 · Filter by author JuncFang-git opened on Dec 31, 2025
  7