Completed
10:11 Full fine-tuned vs. base model unipc
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Wan 2.1 Full Fine-tuning, RES4LYF Clownshark Sampling, and Flex Model Comparisons
Automatically move to the next video in the Classroom when playback concludes
- 1 00:00 Introduction
- 2 00:45 Training
- 3 01:27 Incorrect Epsilons
- 4 01:54 Effect of changes of epsilon value
- 5 04:06 How to change the optimizer settings
- 6 04:42 Weights and biases validation
- 7 05:26 Training configuration
- 8 07:00 Converting diffusers
- 9 07:49 Diffusers keys
- 10 08:47 Conversion dictionary
- 11 09:05 ComfyUI model detection
- 12 09:53 ComfyUI inference
- 13 10:11 Full fine-tuned vs. base model unipc
- 14 10:55 Euler sampler
- 15 12:21 DPMPP_2M
- 16 13:05 DPMPP_2M_SDE_GPU
- 17 13:51 Clownshark Sampler
- 18 14:49 ETA Noise
- 19 15:53 ETA 0 vs. 0.5 vs. 1
- 20 17:53 Sigma Scheduling
- 21 19:22 Sigma schedule plots
- 22 20:20 How to generate no sigmas plot
- 23 21:21 Sigmas schedule comparison
- 24 21:50 Karras version mistake
- 25 22:46 Bong math crash course
- 26 23:33 Bong math qualitative analysis
- 27 24:25 Interesting observation about noise patterns
- 28 24:52 Redone with correct interpolated steps
- 29 25:21 Different seed
- 30 27:00 Wan 2.1 VACE
- 31 28:48 Wan2.1 vs. Flex
- 32 30:30 Male character design
- 33 34:08 Female character design
- 34 37:25 Hard surface design
- 35 38:24 Large scale environment
- 36 40:06 Interior design
- 37 41:37 Wan2.1 refinement
- 38 42:38 Conclusion
