logo
RunComfy
AI Toolkit LoRA Training Guides

Resume AI Toolkit LoRA Training After Preset Steps

Quick guide to resuming LoRA training in AI Toolkit: edit the stopped job, increase Steps to your new total (not extra steps), and restart from the last checkpoint.

Train Diffusion Models with Ostris AI Toolkit

How to continue training after your AI Toolkit job hits the preset Steps (resume the same LoRA)

1) “Is this my problem?”

You set Steps = 2000, started training, and the job stopped automatically once it reached step 2000.

After checking samples / inference, you want more training, but you don’t want to start a brand‑new LoRA from scratch, you want to continue from the existing job/checkpoint.

If that’s you, this guide is for you.

2) What’s happening

In AI Toolkit, Steps = the total training steps for the job.

So when you set Steps = 2000, the trainer did exactly what you asked:

  • train until step 2000
  • then stop

To keep training the same LoRA, you don’t create a new job, you edit the existing job and increase the final total Steps, then run it again. The trainer will resume from the last saved checkpoint and continue.

3) How to resume training (step by step)

Step A — Open the job you want to continue

  1. Go to your Training Queue (or job list)
  2. Find the finished job (the one that stopped at 2000 steps)
  3. Click Edit on that job to open the job editor

Edit button in training job

Step B — Increase the total Steps

  1. Scroll to the Training section
  2. Find Steps
  3. Change it to your new final total step count

Update Steps in training job

Important: This number is the final total, not “extra steps”.

Example:

  • You trained to 2000
  • You want 1000 more
  • Set Steps = 3000 (not 1000)

Step C — Save and restart

  1. Click Update Job
  2. Click Start Training

✅ The job should now continue from the previous run, using the last checkpoint, and train until the new Steps value.

4) If it doesn’t resume (fast checks)

If it starts over from step 0 or refuses to continue, check these:

  • Do you have a checkpoint saved near the end?

    Look at your job outputs / saved files. If your job never saved a checkpoint, it can’t resume.

  • Did you keep checkpoints?

    If your “Max Step Saves to Keep” is low and older saves were pruned, make sure the latest checkpoint still exists.

  • Did you change more than Steps?

    To resume cleanly, keep the same dataset and training settings. Only change Steps.

5) One-line fix summary

To keep training the same LoRA after it stops at your preset Steps:

Edit the finished job → increase Steps to the new final total → Update Job → Start Training.

Ready to start training?