[Feature] Can Lora be cached? When switching workflows whth different Loras

[liuguicen]

Checklist

• 1. If the issue you raised is not a feature but a question, please raise a discussion at https://github.com/mit-han-lab/ComfyUI-nunchaku/discussions/new/choose. Otherwise, it will be closed.
• 2. I will do my best to describe the issue in English.

Motivation

Thank you for the great work. When I switched between different workflows, the two workflows used different Loras, and each time I switched, I needed to load Loras again. Surprisingly, inference only required 3. xs, but loading Loras took and reinference cost 15. xs
Can Lora be cached?

Related resources

2025-07-13 15:30:56.119 - 成功加载图像，形状: torch.Size([1, 614, 440, 3])
0%| | 0/8 [00:00<?, ?it/s]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
12%|█▎ | 1/8 [00:00<00:02, 2.73it/s]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
25%|██▌ | 2/8 [00:00<00:02, 2.73it/s]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
38%|███▊ | 3/8 [00:01<00:01, 2.66it/s]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
50%|█████ | 4/8 [00:01<00:01, 2.68it/s]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
62%|██████▎ | 5/8 [00:01<00:01, 2.68it/s]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
75%|███████▌ | 6/8 [00:02<00:00, 2.64it/s]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
88%|████████▊ | 7/8 [00:02<00:00, 2.65it/s]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
100%|██████████| 8/8 [00:03<00:00, 2.67it/s]
2025-07-13 15:30:59.411 - 开始保存图片
2025-07-13 15:30:59.682 - 保存图片到 D:\aigc\ComfyUI\output\ComfyUI_20250713_153059_541_.webp 成功
Prompt executed in 3.76 seconds
0%| | 0/8 [00:00<?, ?it/s]
Converting LoRAs to nunchaku format: 0%| | 0/57 [00:00<?, ?it/s]
Converting LoRAs to nunchaku format: 11%|█ | 6/57 [00:00<00:00, 56.10it/s]
Converting LoRAs to nunchaku format: 21%|██ | 12/57 [00:00<00:00, 56.11it/s]
Converting LoRAs to nunchaku format: 32%|███▏ | 18/57 [00:00<00:00, 53.54it/s]
Converting LoRAs to nunchaku format: 42%|████▏ | 24/57 [00:00<00:00, 54.76it/s]
Converting LoRAs to nunchaku format: 53%|█████▎ | 30/57 [00:00<00:00, 55.73it/s]
Converting LoRAs to nunchaku format: 63%|██████▎ | 36/57 [00:00<00:00, 56.39it/s]
Converting LoRAs to nunchaku format: 74%|███████▎ | 42/57 [00:00<00:00, 47.66it/s]
Converting LoRAs to nunchaku format: 82%|████████▏ | 47/57 [00:00<00:00, 40.83it/s]
Converting LoRAs to nunchaku format: 91%|█████████ | 52/57 [00:01<00:00, 37.00it/s]
Converting LoRAs to nunchaku format: 100%|██████████| 57/57 [00:01<00:00, 43.27it/s]
[2025-07-13 15:19:16.843] [info] Loading partial weights from pytorch
[2025-07-13 15:19:17.017] [info] Done.
Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
12%|█▎ | 1/8 [00:03<00:21, 3.07s/it]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
25%|██▌ | 2/8 [00:04<00:11, 1.94s/it]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
38%|███▊ | 3/8 [00:05<00:07, 1.57s/it]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
50%|█████ | 4/8 [00:06<00:05, 1.40s/it]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
62%|██████▎ | 5/8 [00:07<00:03, 1.31s/it]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
75%|███████▌ | 6/8 [00:08<00:02, 1.28s/it]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
88%|████████▊ | 7/8 [00:09<00:01, 1.01s/it]Passing txt_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
Passing img_ids 3d torch.Tensor is deprecated.Please remove the batch dimension and pass it as a 2d torch Tensor
100%|██████████| 8/8 [00:09<00:00, 1.22s/it]
2025-07-13 15:19:28.124 - 开始保存图片
2025-07-13 15:19:28.333 - 保存图片到 D:\aigc\ComfyUI\output\ComfyUI_20250713_151928_274_.webp 成功
Prompt executed in 18.86 seconds
got prompt
2025-07-13 15:19:42.962 - 开始从URL加载图像: http://file.baozouptu.intelimeditor.com/release/ml_temp/N4Wd3ukVi.webp
2025-07-13 15:19:43.107 - 读取网络图片成功

[lmxyy]

See https://nunchaku.tech/docs/nunchaku/usage/lora.html#exporting-converted-loras

The current implementation employs single-threaded on-the-fly conversion, which may result in extended processing times, particularly for large LoRA files. We will implement a faster one in the future.

[github-actions]

This issue has been automatically closed due to 30-day inactivity. Please feel free to reopen it with /reopen if needed.